Kenkichi Iwasawa Collected Papers [2001st Edition]
 4431550550,  9784431550556

Table of contents :
Front Matter......Page 1
K. Iwasawa Princeton, 1986......Page 4
First page of Iwasawa's mauscript on cyclotomic fields [62]......Page 6
Contents......Page 8
Preface......Page 12
Introduction......Page 14
Acknowledgements......Page 15
John Coates, lwasawa's work in algebraic number theory......Page 16
[1] Ueber die Struktur der endlichen Gruppen,deren echte Untergruppen sämtlich nilpotent sind, Proc. Phys.-Math. Soc. Japan, 23-1 (1941), 1-4.......Page 27
[2] Über die Einfachheit der speziellen projektiven Gruppen, Proc. Imp. Acad. Japan, 17-3 (1941), 57-59.......Page 31
[3] Über die endlichen Gruppen und die Verbände ihrer Untergruppen, J. Fac. Sci. Tokyo lmp. Univ., Sect I. 4-3 (1941), 171-199.......Page 34
[4] On almost periodic functions, Isô Sûgaku, 4-2 (10/1942), 56-60.......Page 63
[5] On normed rings and a theorem of Segal, I, Zenkoku Shijo Sugaku Danwakai, 246 (12/1942), 1522-1555......Page 70
[6] On the structure of infinite M-groups, Japanese J. Math., 18 (1943), 709-728.......Page 87
[7] On the structure of conditionally complete lattice-groups, Japanese J. Math., 18 (1943), 777-789.......Page 107
[8] Einige Sätze über freie Gruppen, Proc. Imp. Acad. Japan, 19-6 (1943), 272-274.......Page 120
[9] On one-parameter families of probability laws, J. Phys.-Math. Soc. Japan, 17-6 (1943}, 217-220......Page 123
[10] On normed rings and a theorem of Segal, II, Zenkoku Shijo Sugaku Danwakai, 251 (3/1943), 167-186......Page 128
[11] On group rings of topological groups, Proc. Imp. Acad. Japan, 20-2 (1944), 67-70.......Page 140
[12] Über nilpotente topologische Gruppen I, Proc. Japan Acad., 21-3 (1945), 124-137.......Page 144
[13] Der Bezoutsche Satz in zweifach projektiven Räumen, Proc. Japan Acad., 21-4 (1945), 213-222.......Page 158
[14] Zur Theorie der algebraischen Korrespondenzen I, Schnittpunktgruppen von Korrespondenzen, Proc. Japan Acad., 21-4 (1945), 204-212.......Page 168
[15] Zur Theorie der algebraischen Korrespondenzen II, Multiplikation der Korrespondenzen, Proc. Japan Acad., 21-9 (1945), 411-418.......Page 177
[16] On the representation of Lie algebras, Japanese J. Math., 19 (1948), 405-426.......Page 185
[17] On linearly ordered groups, J. Math. Soc. Japan, 1-9 (1948), 1-9.......Page 207
[18] Finite groups and compact groups, Sûgaku, 1-2 (3/1948), 94-95......Page 216
[19] Hilbert's fifth problem, Sûgaku, 1-3 (11/1948), 161-171......Page 219
[20] (with T. Tamagawa) Automorphisms of a function field, Sûgaku, 1-4 (1/1949), 315-316.......Page 231
[21] On some types of topological groups, Ann. of Math., (2) 50 (1949), 507-558.......Page 233
[22] Topological groups with invariant compact neighborhoods of the identity, Ann. of Math., (2) 54 (1951), 345-348.......Page 285
[23] A note on L-functions, "Proc. Int. Congress of Math." (Cambridge, Mass., 1950), Vol.1, Amer. Math. Soc., 1952, p. 322.......Page 289
[24] Some properties of (L)-groups, Proc. Int. Congress of Math. (Cambridge, Mass., 1950), Vol.2, Amer. Math. Soc., 1952, pp. 447-450.......Page 290
[25] (with T. Tamagawa) On the group of automorphisms of a function field, J. Math. Soc. Japan, 3 (1951), 137-147. Corrections, ibid., 4 (1952),100-101, 203-204.......Page 294
[26] On the rings of valuation vectors, Ann. of Math., (2) 57 (1953), 331-356.......Page 309
[27] On solvable extensions of algebraic number fields, Ann. of Math., (2) 58 (1953), 548-572.......Page 335
[28] A note on Kummer extensions, J. Math. Soc. Japan, 5 (1953), 253-262.......Page 360
[29] On Galois groups of local fields, Trans. Amer. Math. Soc., 80 (1955), 448-469.......Page 370
[30] Galois groups acting on the multiplicative groups of local fields, "Proc. Int. Sympos. on Algebraic Number Theory" (Tokyo and Nikko, 1955), Science Council of Japan, Tokyo, 1956, pp. 63-64.......Page 392
[31] A note on the group of units of an algebraic number field, J. Math. Pure and Appl., (9) 35 (1956), 189-192.......Page 394
[32] A note on class numbers of algebraic number fields, Abh. Math. Sem. Univ. Hamburg, 20 (1956), 257-258.......Page 398
[33] On some invariants of cyclotomic fields, Amer. J. Math., 80 (1958), 773-783. Erratum, ibid., 81 (1959), 280.......Page 400
[34] Sheaves for algebraic number fields, Ann. of Math., (2) 69 (1959), 408-413.......Page 412
[35] On \Gamma-extensions of algebraic number fields, Bull. Amer. Math. Soc., 65 (1959), 183-226.......Page 418
[36] On some properties of \Gamma-finite modules, Ann. of Math., (2) 70 (1959), 291-312.......Page 462
[37] On the theory of cyclotomic fields, Ann. of Math., (2) 70 (1959), 530-561.......Page 484
[38] On local cyclotomic fields, J. Math. Soc. Japan, 12 (1960), 16-21.......Page 516
[39] A class number formula for cyclotomic fields, Ann. of Math., (2) 76 (1962), 171-179.......Page 522
[40] On a certain analogy between algebraic number fields and function fields, Sûgaku, 15-2 (10/1963), 65-67......Page 531
[41] On some modules in the theory of cyclotomic fields, J. Math. Soc. Japan, 16 (1964), 42-82.......Page 535
[42] Some results in the theory of cyclotomic fields, Proc. Sympos. Pure Math., Vol. VIII, Amer. Math. Soc., 1965, pp. 66-69.......Page 576
[43] (with C. C. Sims) Computation of invariants in the theory of cyclotomic fields, J. Math. Soc. Japan, 18 (1966), 86-96.......Page 580
[44] A note on ideal class groups, Nagoya Math J., 27 (1966), 239-247.......Page 591
[45] Some modules in local cyclotomic fields, "Les Tendances Géom. en Algèbre et Théorie des Nombres", Éditions du Centre National de la Recherche Scientifique, Paris,1966, pp. 87-96.......Page 600
[46] On explicit formulas for the norm residue symbol, J. Math. Soc. Japan, 20 (1968), 151-165.......Page 610
[47] Analogies between number fields and function fields, "Some Recent Advances in the Basic Sciences, Proc. Annual Sci. Conf." (New York, 1965-1966), Vol. 2, Belfer Grad. School Sci.,Yeshiva Univ., New York, 1969, pp. 203-208.......Page 625
[48] On p-adic L-functions, Ann. of Math., (2) 89 (1969), 198-205.......Page 631
[49] On some infinite Abelian extensions of algebraic number fields, "Actes du Congrés Int. Math." (Nice, 1970), Tome 1, Gauthier-Villars, Paris, 1971, pp. 391-394.......Page 639
[50] Skew-symmetric forms for number fields, Proc. Sympos. Pure Math., Vol. XX, Amer. Math. Soc., 1971, p. 86.......Page 643
[51] On the \mu-invariants of cyclotomic fields, Acta Arith., 21 (1972), 99-101.......Page 644
[52] On Z_$\ell$-extensions of algebraic number fields, Ann. of Math., (2) 98 (1973), 246-326.......Page 647
[53] On the \mu-invariants of Z_$ell$-extensions, "Number Theory, Algebraic Geometry and Commutative Algebra, in honor of Yasuo Akizuki", Kinokuniya, Tokyo, 1973, pp. 1-11.......Page 728
[54] A note on Jacobi sums, "Convegno di Strutture in Corpi Algebrici" (Rome, 1973), INDAM, Symposia Mathematica, Vol, XV, Academic Press, London, 1975, pp. 447-459.......Page 739
[55] A note on cyclotomic fields, Invent. math., 36 (1976), 115-123.......Page 752
[56] Some remarks on Hecke characters, "Algebraic Number Theory" (Kyoto, 1976), Japan Soc. Promotion Sci., Tokyo, 1977, pp. 99-108.......Page 761
[57] On p-adic representations associated with Zp-extensions, "Automorphic Forms, Representation Theory and Arithmetic" (Bombay, 1979), Tata Inst. Fund. Res., Bombay, 1981, pp. 141-153.......Page 771
[58] Riemann-Hurwitz formula and p-adic Galois representations for number fields, Tôhoku Math. J., (2) 33 (1981), 263-288.......Page 784
[59] On cohomology groups of Zp-extensions, "Zp-extensions and Related Topics" (Kyoto, 6/1981; G. Fujisaki, ed.), Kokyuroku, No. 440, RIMS, Kyoto Univ., 1981, pp. 76-84......Page 810
[60] On cohomology groups of units for Zp-extensions, Amer. J. Math., 105 (1983), 189-200; "Geometry and Number Theory, A Volume in Honor of André Weil" (J.-P. Serre and G. Shimura eds.), The Johns Hopkins University Press, 1983, pp. 189-200.......Page 815
[61] A simple remark on Leopoldt's conjecture, "Report of the Conference on Algebraic Number theory" (RIMS, Kyoto, 7/1984; K. Shiratani, ed.), 1984, pp. 45-54......Page 827
[63] A note on capitulation problem for number fields, Proc. Japan Acad., 65-2, Ser. A (1989), 59-61.......Page 838
[64] A note on capitulation problem for number fields. II, Proc. Japan Acad., 65-6, Ser. A (1989), 183-186.......Page 841
[65] On papers of Takagi in number theory, "Teiji Takagi Collected Papers", Springer-Verlag Tokyo, 1990, Appendices, I, pp. 342-351.......Page 845
[66] A letter to J. Dieudonné, "Zeta Functions in Geometry"(Tokyo, 1990), Adv. St. in Pure Math., 21, 1992, pp. 445-450.......Page 855
[U1] A note on unitary representations, (lectured in a Colloquium at Kyoto Univ., Dec. 1987).......Page 861
[U2] Automorphisms of Galois groups over number fields, (lectured in the "Symposium on L-functions and Galois properties of algebraic number fields",
the Univ. of Durham, Sept. 1975).......Page 865
[U3] Some problems on cyclotomic fields and Zp-extensions, (lectured in "Arithmetik der ABELschen Zahlkörper und Klassenkörper der komplexen Multiplikation",
Math. Forschungsinstitut Oberwolfach, March 1979).......Page 871
[U4] A simple remark on Leopoldt's conjecture, (lectured in the "Conference on
Algebraic Number Theory", Res. Inst. Math. Sci., Kyoto Univ., July 1984).......Page 880
[U5] On integral representations of some finite groups, (lectured in Komaba seminar at the Univ. of Tokyo, June 1989).......Page 889
Bibliography......Page 893
Back Cover......Page 899
[A1] On groups and the associated lattices (909. 群トソノ lattice ニツイテ), ZSSD, 211(19410315), 76-81......Page 900
[A2] On groups and the associated lattices II (群とその lattice についてII), Iso
Sugaku (位相数学), 4-1 (2/1942), 47-50......Page 906
[A3] On groups and the associated lattices III (975. 群トソノ lattice ニツイテ III), ZSSD, 225(19411027), 519-542......Page 910
[A4] A remark on lattice groups (1029. 束群ニ関スル一注意), ZSSD, 234(19420323), 912-914......Page 934
[A5] On lattice groups (1044. 束群ニツイテ), ZSSD, 235(19420416), 1030-1048......Page 937
[A6] On a generalization of central series (1089. 核心群列ノ一拡張ニツイテ), ZSSD, 246(19421214), 1555-1563......Page 956
[A7] On the correspondence between groups and ideals in the group rings (1090. 群ト群環ニ於ケル Ideal トノ対応ニ就イテ), ZSSD, 246(19421214), 1563-1591......Page 965
[A8] A remark on solvable groups (1149. 可解群ニ関スル一注意), ZSSD, 258(19431025), 539-544......Page 994
[A9] An example of a group (1168. 群ノ一例), ZSSD, 262(19440320), 67-69......Page 1000
[A10] On faithful representations of a Lie algebra (23. Lie 環ノ同型表現ニツイテ), ZSSD (Ser.2), 3(19470220), 1-6......Page 1003
[A11] Review of "Theory of differential forms - Grassmann algebras and Lie groups (微分式論―グラスマン代数とリー群)" by Yukiyoshi Kawada (Kawade Shobo(河出書房), Tokyo,1951), Math. Reviews 14 (1953), 410......Page 1009
[A12] Review of "Allgemeine Theorie der algebraischen Zahlen" by Philipp Furtwängler (B. G. Teubner Verlagsgesellschaft, Leipzig, 1953), Math. Reviews 15 (1954), 404......Page 1010
[A13] Review of "Introduction to algebraic number theory" by Henry B. Mann (The Ohio State Univ. Press, Columbus, Ohio, 1955), Math. Reviews 17 (1956), 240-241......Page 1011
[A14] Review of "Topological transformation groups" by Deane Montgomery and Leo Zippin (Interscience Publishers, New York -London, 1955), Math. Reviews 17 (1956), 383-384......Page 1013
[A15] Review of Die Berechnung der Klasssenzahl Abelscher Körper über quadratischen Zahlkörpern by Curt Meyer (Akademie-Verlag, Berlin, 1957), Bull. Amer. Math. Soc., 64 (1958), 211-212......Page 1015
[A16] Impressions of the International Congress of Mathematicians, I (Congress印象記I), Sugaku (数学) 14-3 (2/1963), 174-175......Page 1017
[A17] Review of "Introduction to quadratic forms" by O. T. O'Meara (Academic Press, Inc., Publishers, New York; Springer-Verlag, Berlin-Göttingen-Heidelberg, 1963), Math. Reviews 27 (1964), 482-483......Page 1019
[A18] On the mathematical works of Prof. Hiraku Toyama - around the thesis "Non-abelian theory of algebraic functions" (遠山啓教授の数学的業績ー学位論文「代数関数の非アーベル的理論」を中心に), Sugaku Seminar (数学セミナー) (3/1980), 24-28......Page 1021
[A19] 120 minutes with Prof. Kenkichi Iwasawa (Interview by the editors of Sugaku)(岩澤健吉先生のお話を伺った120分(編集部)), Sugaku(数学), 45-4 (101990), 366-372......Page 1026
Round-table talk with 3 overseas Japanese mathematicians (座談会・在外数学者の生活と意見), Sugaku Seminar (数学セミナー) (2/1963), 4-12......Page 1033

Citation preview

s c i t a m e h t a nM si k r o dW e t c e l l o rC e g n i r p S

1

a w a s a w il h c i k n e K

s r e p a dP e t c e l l o C

~Springer

Kenkichi Iwasawa

Collected Papers Editor in Chief lchiro Satake

Editors Genjiro Fujisaki • Kazuya Kato• Masato Kurihara• Shoichi Nakajima

Reprint of the 2001 Edition

~ Springer

Author Kenkichi Iwasawa (1917-1998) Princeton University Princeton, NJ, USA

Editor in Chief Ichiro Satake (emeritus) University of California, Berkeley Berkeley, CA, USA and Tohoku University Sendai, Japan

Editors Genjiro Fujisaki (emeritus) The University of Tokyo Tokyo, Japan

KazuyaKato University of Chicago Chicago, IL, USA

Masato Kurihara Keio University Yokohama, Japan

Shoichi Nakajima Gakushuin University Tokyo, Japan

Contributor John Coates University of Cambridge Cambridge, United Kingdom

ISSN 2194-9875 ISBN 978-4-431-55055-6 Springer Tokyo Heidelberg New York Dordrecht London Library of Congress Control Number: 2014957321

© Springer-Verlag Tokyo 2001, Reprint 2014 This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed. Exempted from this legal reservation are brief excerpts in connection with reviews or scholarly analysis or material supplied specifically for the purpose of being entered and executed on a computer system, for exclusive use by the purchaser of the work. Duplication of this publication or parts thereof is permitted only under the provisions of the Copyright Law of the Publisher's location, in its current version, and permission for use must always be obtained from Springer. Permissions for use may be obtained through RightsLink at the Copyright Clearance Center. Violations are liable to prosecution under the respective Copyright Law. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. While the advice and information in this book are believed to be true and accurate at the date of publication, neither the authors nor the editors nor the publisher can accept any legal responsibility for any errors or omissions that may be made. The publisher makes no warranty, express or implied, with respect to the material contained herein. Printed on acid-free paper Springer is part of Springer Science+Business Media (www.springer.com)

K. lwasawa

Princeton, 1986

し ー

IIIII T I J : ・ : : : c : : 」 _ _ _J _ 」 」

—_____.J..

ーニエFり召丑五[il~J~~::~ い [ り . . !..~堕 」

工 □II工I~.r.~~IエIJりiJ~,1冦[iT『且土止し占~"'b I . _ 」

1 , . -

F i r s tp a g eo fI w a s a w a ' sm皿 u s c r i p tonc y c l o t o m i cf i e l d s[ 6 2 ]

Contents

xi

Preface Introduction

xiii

Acknowledgements

xiv

Iwasawa's work in algebraic number theory (by J. Coates)

xv

[1] Ueber die Struktur der endlichen Gruppen, deren echte Untergruppen 1

samtlich nilpotent sind [2] Ober die Einfachheit der speziellen projektiven Gruppen

5

[3] Ober die endlichen Gruppen und die Verbiinde ihrer Untergruppen

·8

[4] On almost periodic functions (in Japanese)

37

[5] On normed rings and a theorem of Segal, I (in Japanese)

44

[6] On the structure of infinite M-groups

61

[7] On the structure of conditionally complete lattice-groups

81

[8] Einige Satze fiber freie Gruppen

94

[9] On one-parameter families of probability laws (in Japanese)

97

[10] On normed rings and a theorem of Segal, II (in Japanese)

102

[11] On group rings of topological groups

114

[12] Ober nilpotente topologische Gruppen I

118

· [13] Der Bezoutsche Satz in zweifach projektiven Raumen

132

[14] Zur Theorie der algebraischen Korrespondenzen I, Schnittpunktgruppen von Korrespondenzen

142

[15] Zur Theorie der algebraischen Korresj:)ondenzen II, Multiplikation der Korrespondenzen

151

[16] On the representation of Lie algebras

159

[17) On linearly ordered groups

181

[ 18) Finite groups and compact groups (in Japanese)

190

[19] Hilbert's fifth problem (in Japanese)

193

[20] (with T. Tamagawa)Automorphisms of a function field (in Japanese)

205

[21] On some types of topological groups

207

vii

viii

Contents

[22) Topological groups with invariant compact neighborhoods of the identity

259

[23] A note on L-functions

263

[24) Some properties of (L)-groups

264

[25) (with T. Tamagawa) On the group of automorphisms of a function field

268

[26) On the rings of valuation vectors

283

[27) On solvable extensions of algebraic number fields

309

[28) A note on Kummer extensions

334

[29) On Galois groups oflocal fields

344

[30] Galois groups acting on the multiplicative groups of local fields

366

[31] A note on the group of units of an algebraic number field

368

[32) A note on class numbers ofalgebraic number fields

372

[33) On some invariants of cyclotomic fields

374

[34) Sheaves for algebraic number fields

386

[35) On f-extensions of algebraic number fields

392

[36] On some properties off-finite modules

436

[37) On the theory of cyclotomic fields

458

[38) On local cyclotomic fields

490

[39) A class number formula for cyclotomic fields

496

[40] On a certain analogy between algebraic number fields and function fields (in Japanese)

505

[41] On some modules in the theory of cyclotomic fields

509

[42) Some results in the theory of cyclotomic fields

550

[43) (with C. C. Sims) Computation of invariants in the theory of cyclotomic fields

554

[44) A note on ideal class groups

565

[45) Some modules in local cyclotomic fields

574

[46) On explicit formulas for the norm residue symbol

584

[47) Analogies between number fields and function fields

599

[48] On p-adic L-functions

605

[49) On some infmiteAbelian extensions of algebraic number fields

613

[50) Skew-symmetric forms for number fields

617

Contents

ix

[51] On the µ-invariants of cyclotomic fields

618

[52] On Zt·extensions of algebraic number fields

621

[53] On theµ-invariants ofZrextensions

702

[54] A note on Jacobi sums

713

[55] A note on cyclotomic fields

726

[56] Some remarks on Hecke characters

735

[57] On p-adic representations associated with Zp-extensions

745

[58] Riemann-Hurwitz formula and p-adic Galois representations for number

fields

758

[59] On cohomology groups ofZp-extensions (in Japanese)

784

[60] On cohomology groups of units for Zp-extensions

789

[61] A simple remark on Leopoldt's conjecture (in Japanese)

801

[62] Some problems on cyclotornic fields (in Japanese)

805

[63] A note on capitulation problem for number fields

812

[64] A note on capitulation problem for number fields. II

815

[65] On papers of Takagi in number theory

819

[66] A letter to J. Dieudonne

829

[Ul] A note on unitary representations

835

[U2] Automorphisms of Galois groups over number fields

839

[U3] Some problems on cyclotmnic fields and Zp·extensions

845

[U4] A simple remark on Leopoldt's conjecture

854

[US] On integral representations of some finite groups

863

Bibliography

867

Preface

Kenkichi Iwasawa was born on September 11, 1917 in Shinshuku-mura (Kiryuushi) in Gunma prefecture. After elementary school he was educated in Tokyo, where he first attended Musashi High School, one of the few special high schools with seven grades (which combine four years of middle school education with three years of high school education). In 1937 he entered the University of Tokyo (then Tokyo Imperial University) to do undergraduate studies in mathematics. After graduation in 1940, he continued his graduate studies there, became Assistant and then Instructor, and obtained the degree of Doctor of Science in 1945 (this was during the World War II, 1941-1945). That year he became seriously ill with pleurisy and could return to his post only in April 1947. He was appointed Assistant Professor at the University of Tokyo in 1949. In the summer of 1950, Iwasawa was invited to the International Congress of Mathematicians held in Cambridge, Massachusetts to give two talks ([23], [24)), and then spent two academic years as a member at the Institute for Advanced Study in Princeton. While preparing to return to Tokyo in the spring of 1952, he received an offer from the Massachusetts .Institute of Technology, which he accepted; he was Assistant Professor there from 1952, Associate Professor from 1955, and Professor from 1958. In 1966 he moved to Princeton and was Professor at Princeton University from 1966 and H. B. Fine Professor of Mathematics from 1978 until his retirement in 1986. Meanwhile, he was also a member at the Institute for Advanced Study for 1957-58, 1966-67, Spring of 1971, Spring of 1979, and Spring of 1982.* lwasawa and his wife returned to live in Tokyo in the fall of 1987. Though he no longer did university teaching, he regularly attended seminars at the University of Tokyo, which was a great inspiration to the younger generation in Japan. He died of acute pneumonia on October 26, 1998 at the age of 81. lwasawa was awarded the Asahi Cultural Prize in 1959, the Prize of the Japan Academy in 1962, the Frank Nelson Cole Prize in Algebra and in Number Theory of the American Mathematical Society in 1962, and the Fujiwara Prize in 1979. Iwasawa made a number of fundamental contributions in group theory and in algebraic number theory; the main body of the latter, now known as lwasawa theory, has many applications in a wider area of arithmetic algebraic geometry (including the solution of Fermat's Last Theorem). We are very fortunate to be able to include an expository article on lwasawa theory by John Coates, to whom we are very grateful for his invaluable contribution to this volume. Thanks are due to Springer-Verlag Tokyo for bringing out this publication, which, we hope, will enable future mathematicians to inherit the magnificent * These data on the academic career of K. Iwasawa were kindly provided by the Faculty of Sciences, the University of Tokyo and the Department of Mathematics, Princeton University. xi

xii

Preface

ideas and results of Kenkichi Iwasawa, one of the most original and influential mathematicians of our time.

April, 2001

Editorial Committee G. Fujisaki, K. Kato, M. Kurihara, S. Nakajima, and I. Satake (Chair)

Introduction

These two volumes contain all 66 published papers of Kenkichi Iwasawa ([1](661), of which 47 a.re in English, 8 in German, and 11 in Japanese; the secCiJ.id volume contains also 5 unpublished papers in English ([Ul)-[U5]). For each of these 11 papers in Japanese and 5 unpublished papers, either an (English) abstract or notes by the editors or other specialists are provided. Those papers in Japanese ([5], [10], [59J, [61], [62]) which were originally photocopies of hand~ written manuscripts were reproduced here from newly typed InE;X files. For a fow papers, whenever it seemed appropriate, the editors added some comments or corrections of misprints (indicated by asterisks). Most of the reprints and manuscripts were provided by the late Mrs. Aiko Iwasawa, to whom the editors are most grateful. We note that there enst further some 19 articles by K. Iwasawa ([Al]-[Al9]), including 10 papers in Japanese published in Zenkoku Shijo Sugaku Danwakai and Iso Sugalcu (before 1945) and 6 book reviews, not included in these volumes. The bibliography at the end of each volume gives the list of all publications of K. Iwasawa, including 5 books and lecture notes ([B1HB5]) as well as 19 articl~ mentioned above which are not reproduced here. We note that the numbering of the papers in this list from [1] to [62] coincides with that giv~n in the following volume, published in commemoration of his seventieth birthday: J. Col:l.tes, R. Greenberg, B. Mazur, l:l.nd I. Sa.take (eds.), Algebraic Number Theory - in honor ofKenkichi Iwasawa., Adv. St. in Pure Math., VoL 17, Kinokuniya, Tokyo and Acad. Press, 1989.

xiii

Acknowledgements

The editors would like to thank the original publishers of Kenkichi lwasawa's papers for granting permission to reprint them here. The numbers following each. publisher correspond to the numbering of the articles in the table of contents.

Academic Press Ltd. London, UK: [54] The American Mathematical Society: (23], (24], (29], (35], (42], (50] Centre National de la RechercheScientifique: {45] Department of Mathematics, Graduate School of Science, Osaka University: [5], [10] Gauthier-Villars - Editions Scientifiques et Medicales Elsevier: [31] Graduate School of Mathematical Sciences, University of Tokyo: (3] Graduate School of Mathematics, Nagoya University: [44] Institute of Mathematics, Polish Academy of Sciences: [51] The Japan Academy: [2], (8], [11], (12], (13], [14], (15], (63], (64] Japan Society for the .Promotion of Science: [56] The Johns Hopkins University Press: [33], [60] Kinokuniya Company Ltd;: (53] The Mathematical Society of Japan: [l], (6], (7], (9], [16], (17], [18], (19], (20], (25], [28],

[38], (40], (41], (43], [46], (66] Mathematical· Institute, Tohoku University: (58] Princeton University Press: [21], [22], (26], [27], (34], [36], [37], [39], (48], [52) Research Institute for Mathematical Sciences, Kyoto University: [59], [61], [62] The Science Council of Japan: (30] Springer-Verlag: [55], (65] Tata Institute of Fundamental Research: [57] Vandenhoeck & Ruprecht: (32] Yeshiva University: [47]

xiv

lwasawa's work in algebraic number theory John Coates Emmanuel College, Cambridge CB2 3AP, England

A brief glance at the articles in these volumes will quickly convince the reader that Iwasawa wrote mathematics, as indeed he also lectured it, with beauty, elegance, and precision. Nothing is a substitute for reading his original papers, and the aim of the present article is merely to provide, with the benefit of hindsight, a short and inevitably personal overview of the main themes of Iwasawa's work in algebraic number theory, together with a few comments on the much wider circle of problems in arithmetical algebraic geometry to which his ideas have been fruitfully applied. I have concentrated on his work on cyclotomic fields and Zp-extensions of number fields (these are essentially the papers from {26} onwards in the present volumes). t However, it should not be forgotten that Iwasawa also made important contributions to other areas of mathematics. Most notable amongst these are his celebrated paper {21} on locally compact groups and Hilbert's fifth problem, and his brief note {23} {see also {66}) in which, independently of J. Tate, he also discovered the adelic approach to Hecke's £-functions, which has so deeply influenced the modern theory of automorphic forms. Finally, we also urge the reader to consult the excellent and much more detailed article on Iwasawa's work by R. Greenberg [12]. Let F be a finite extension of the rational field Q. Classical algebraic number theory seeks to describe the arithmetic of F (the ideal class group, the unit group, the set of rational primes which split completely in F, the Galois group of the maximal abelian extension of F which is unramified outside a given finite set of places, ... ). Beginning with the work of Dirichlet, Dedekind, and Kummer, it has long been known that there is more than one connexion between these arithmetic problems and the complex zeta and £-functions attached to F. The new idea underlying Iwasawa's work was that one could obtain important arithmetic invariants of F, and establish deep new relations between the arithmetic of F and zeta values, by studying certain infinite towers of number fields over F, which he called Zp-extensions. Here p denotes an arbitrary prime number, and its choice gives a p-adic flavour to all of Iwasawa's work. By definition, a Zp-extension of F is an infinite tower of fields

(1)

F

= Fo C

F1

C ··· C

Fn

C ... ,

where, for each n ;;:, 0, Fn is a cyclic extension of F of degree pn. The reason for this terminology is that if we define F00 to be the union of all the Fn(n ;;:, 0), then F 00 is a Galois extension of F whose Galois group over F, which we shall always t I use square brackets [1], ... to denote the references at the end of this article, and curly brackets {26}, ... to denote the references to Iwasawa's papers as numbered in these volumes.

xv

xvi

John Coates

denote by r, is topologically isomorphic to the additive group of p-adic integers Zp. It is easy to see that every F has a unique Zp-extension contained in F(µp"° ), where µpoo denotes the group of all p-power roots of unity; this Zp-extension is called the cyclotomic Zp-extension of F. However, if F is not totally real, class field theory shows that F admits infinitely many Zp-extensions distinct from the cyclotomic one. Some of these non-cyclotomic Zv-extensions also turn out to be very useful for the study of arithmetic questions. Of course, Zp-extensions are the simplest examples of Galois extensions of F whose Galois group is a p-adic Lie group of positive dimension, and it is a fascinating question to speculate on how much of Iwasawa's work might eventually be extended to more general p-adic Lie extensions of F. There is one other important notion which underlies much of Iwasawa's work. Let G denote an arbitrary profinite group (we shall mainly be interested in the case G = r, but the remarks we shall make hold in complete generality). The Iwasawa algebra A( G) of G is defined by (2)

A(G)

= limZp[G/U],

u

where U runs over all open normal subgroups of G, and Zp[G/U] denotes the ordinary group ring of the finite group G /U with coefficients in Zp. The beauty of the Iwasawa algebra A(G) is that it has both an algebraic and an analytic interpretation. The algebraic avatar of A( G) is the fact that we can naturally extend the continuous action of G on any compact Zp-module X to an action of the whole Iwasawa algebra A(G). This idea is explicit in Iwasawa's first papers {35}, {36} on Zp-extensions, but J-P. Serre (see [27], (28]) was the first to stress how one could exploit it more fully, noting in particular that, when G = Zp d for some integer d ~ 1, then A(G) could be identified with the ring Zp[[T1, ... , Td]J of formal power series in d variables, with coefficients in Zp. The analytic avatar of A( G) is the fact that it can be identified with the algebra of measures on G with values in Zp, thus allowing us to define the integral

fatdµ for all µ in A( G) and all continuous functions f from G to Qp. Once again, this idea first appears implicitly in Iwasawa's work (see his paper {48} giving a new construction of the Kubota-Leopoldt p-adic £-functions). It was later systematically developed by B. Mazur (see (201). Perhaps Iwasawa's most remarkable discovery is the fact that, at least in some important cases, there is a similar deep algebraic and analytic dichotomy in the arithmetic of Zp-extensions. A precise formulation of this dichotomy is often called "the main conjecture".

Algebraic Theory Roughly speaking, the algebraic side of Iwasawa theory refers to the methods and results which do not involve. making use of the special values of zeta and £-functions. Let F he a finite extension of Q, and let F 00 denote an arbitrary Zp-extension of F. As above, we write r for the Galois group of Foo over F, and A(r) for the Iwasawa algebra of r. Compact A(r)-modules arise naturally

Iwasawa's work in algebraic number theory

xvii

in the theory of Zp-extensions as follows. Let N 00 denote any abelian p-extension of F00 , which is Galois over the base field F (typical examples are when N 00 is the maximal unramified abelian p-extension of Foo, or the maximal abelian p-extension of F 00 which is unramified outside the primes of F 00 lying above p). Write G(N00 / F 00 ) for the Galois group of N 00 which is clearly a compact Zpmodule. The key point is that G(N00 /F00 ) has a natural continuous action of r given by

X= axa-l where a denotes any lifting of a to the Galois

a•

(3)

for a in rand x in G(N00 /F00 ), group of N 00 over F. But then, as remarked above, this r-action can be extended to a continuous action of the whole lwasawa algebra A(r) on G(N00 / F 00 ). Much of the algebraic content of Iwasawa theory is concerned with the study of such Galois groups G(N00 / F00 ) as A(r)-modules. It is usually easy to show that G(N00 / F00 ) is finitely generated over A(r), but much more delicate to determine its rank as a A(r)-module. Iwasawa's first general algebraic result, which is valid for an arbitrary Zpextension F 00 over F, was the following asymptotic formula proven in {35}. Let An denote the p-primary subgroup of the ideal class group of Fn, and write pe,. for the order of An. Then there exist integers >., µ, and v, depending only on the Zp-extension F 00 / F, such that, for all sufficiently large n, we have (4)

en

= >.n + µpn + v.

Two basic facts lie behind the proof of (4). The first is that G(L 00 / Foe,), where L 00 denotes the maximal unramified abelian p-extension of F =, is a finitely generated torsion A(r)-module. The second is that there exists a structure theory for finitely generated torsion modules over A(r), which is only marginally different from the structure theory for such modules over a principal ideal domain. In {35}, Iwasawa gave an ad hoc proof of this structure theory, but J-P. Serre 128] pointed out that it followed from known results in commutative algebra, since A(r) is isomorphic to the ring Zp[[T]J of formal power series in an indeterminate T with coefficients in Zw Iwasawa himself has said that the discovery of the asymptotic formula (4) suggested to him that there might be deep analogies between the p-primary subgroup of the ideal class group of a Zp-extension F00 of F and the p-primary subgroup of the group of points on the Jacobian variety of a curve over a finite field. This idea seems to have motivated much of his subsequent work. Of course, the existence of certain parallels between the arithmetic of number fields and curves over finite fields had been observed long before this, but Iwasawa's insights were to breathe new life into this old theme. Inspired by this analogy, he conjectured that the invariantµ appearing in (4) should always be zero when F00 is the cyclotomic Zp-extension of F. This conjecture was proved by B. Ferrero and L. Washington [8} (and a rather different proof was given later by W. Sinnott 130]) when F is an abelian extension of Q, but it remains open in general. In the early 1970's, Iwasawa {53} discovered the first examples of non-cyclotomic Zpextensions whereµ > 0. Today, there remain many interesting open questions about Iwasawa's invariants>. andµ, and their possible generalizations. We only mention three questions, all of which seem rather difficult. Firstly, if F is totally

:xviii

John Coates

real and F00 / Fis the cyclotomic Zp-extension, R. Greenberg [10] has conjectured that we always have ..\ = µ = 0 (this is equivalent to the assertion that when F is totally real, the maximal unramified abelian p-extension of F 00 is always a finite extension of F 00 ). Secondly, if we take F00 to be the cyclomotic Zp-extension of F and vary p over all primes, one can ask whether ..\ is bounded, by analogy with the Jacobian variety of a curve over a finite field (see [12])? Thirdly, one can ask if there is some generalization of (4) to arbitrary pro-p ,1radic Lie extensions of F. Let K 00 denote a Galois extension of F, whose Galois group G is a pro-p p-adic Lie group of positive dimension d. For each integer n ;;i: 1, let Pn(G) denote the n-th subgroup of the lower central p-series of G, and define Kn to be the fixed field of Pn(G). Again, we write pe,. for the order of the p-primary subgroup of the ideal class group of Kn. Do there exist non-negative integers ..\ and µ such that (5)

as n - oo? When G = Zp d, the formula (5) is true and was proven by A. Cuoco and P. Monsky [6]. Probably the most important of Iwasawa's subsequent papers on the algebraic side of his theory is {52}, which contains much material from his courses at Princeton. Let F 00 be an arbitrary Zp-extension of F. The first six sections of {52} carry out a systematic study of G(L 00 / F00 ) as a A(r)-module, where L 00 again denotes the maximal unramified abelian p-extension of F 00 • One surprising new result proven there is that, if An denotes the p-primary subgroup of the ideal class group of Fn, then for all m ;;i: n ;;i: 0, the order of the kernel of the natural map from An to Am is bounded independent of n and m. While this kernel is known to be non-zero in general (see [10]), it is still unknown if this can happen when F = Q(µp) and F 00 = Q(µp"") for some odd prime p; here µp denotes the group of p-th roots of unity, and µp"" the group of all p-power roots of unity. The next two sections of {52} specialize to the case of the cyclotomic Zp-extension F 00 / F. Let M 00 denote the maximal abelian p-extension of F00 which is unramified outside p. Jwasawa then determines the exact A(r)-rank of G(M00 /F00 ), a result which turns out to be equivalent to establishing a weak form of H. Leopoldt's conjecture on the p-adic independence of the global units of Fn as n --, oo. He also proves that G(M00 / F00 ) has no non-zero finite A(r)submodule. The final part of {52} gives a detailed construction in the number field case of a skew-symmetric bilinear form, which is an analogue of the Weil pairing on the Tate module of the Jacobian of a curve over a finite field.

Analytic Theory I believe that the most remarkable of all Iwasawa's discoveries on Zp-extensions was his work on what became known later as the "main conjecture". This work provided the key to our complete understanding of the connexion, partially discovered by Kummer, between the arithmetic of the field Q(µp"") and the values of the zeta function at the odd negative integers. Today we have become so used to almost routinely formulating (but, alas, very rarely proving) "main conjectures" for elliptic curves, modular forms, motives, ... , that it is easy to forget how much difficulty Iwasawa had in ·uncovering even the statement of his "main conjecture".

lwasawa's work in algebraic number theory

xix

The interested reader can see the evolution of his ideas at first hand in his papers {39}, {41}, {42}, {45}, {46}, {47}, {48}, especially {41} and {48}. We now discuss Iwasawa's work on the "main conjecture" in some detail, and with the benefit of hindsight. In connexion with his work on Fermat's Last Theorem, Kummer defined an odd prime p to be irregular if p divides the class number of the field Q(µp), Kummer himself proved a mysterious and unexpected connexion between the irregularity of p and the special values of the Riemann zeta function ((s). Fork= 2,4,6, ... , we have ((1- k) = -Bk/k, where the Bk are the Bernoulli numbers defined by the expression

Kummer proved that pis irregular if and only if p divides the numerator of at least one of the rational numbers ((1 - k), with k = 2, 4, ... ,p - 3. Let K 00 be the maximal real subfield of Q(µp"" ), that is

and let G be the Galois group of K 00 over Q (the structure of G is given by G = r x A, where r is isomorphic to Zp, and A is cyclic of order (p - 1) /2). Let M 00 be the maximal abelian p-extension of K 00 which is unramified outside p, and put (6)

Now the exact analogue of (3) defines an action of G on X 00 , which we can then extend to an action of the whole Iwasawa algebra A(G). Similar algebraic arguments to those mentioned earlier then show that X 00 is finitely generated over A(G), and has A(G)-rank equal to O (in the sense that X 00 is annihilated by an element of A(G) which is not a divisor of zero). But there is a simple structure theory for finitely generated A( G)-modules, coming essentially from the structure theory of A(f)-modules, and the fact that G = A x r mentioned above. It follows that there exist an integer Tp ;i; 1, and elements f1, ... ,f-r,, of A(G), which are not divisors of zero such that we have an exact sequence of A( G)-modules Tp

(7)

0-. X 00

-.

E9A(G)/fiA(G)-. D-. 0, i=l

where D is a A( G)-module of finite cardinality. It is not obvious that there is a connexion between the module X 00 and the irregularity of a prime number p, but in fact it can be shown by algebraic arguments that p is irregular if and only if X 00 =/=- O. Thus, in view of Kummer's criterion for the irregularity of p, it follows that X= =/=- 0 if and only if p divides at least one of the special values ((1- k), with k = 2, 4, ... ,p- 3. What is the connexion between this mysterious statement and the exact sequence (7) of A( G)-modules? To formulate the answer, Iwasawa realized that one had to use the analytic avatar of A(G). Earlier, T. Kubota and H. Leopoldt [17J, using congruences going back to Kummer, had proven the existence of a p-adic analogue of ((s), but their construction gave no hint of any connexion with A( G). Motivated by

rea:

John Coates

arithmetic arguments emerging from explicit reciprocity laws in {41 }, Iwasawa had been led to study the classical Stickelberger elements in the tower of number fields Q(µp""). In a beautifully simple paper {48}, Iwasawa brought these two lines of thought together, and proved the following result, modulo a slight change of language. We say that an element r.p of the ring of fractions of A( G) is a pseudomeasure if (O' - l)r.p belongs to A(G) for all O' in G (intuitively, one should think of such a r.p as possibly having a simple pole at the trivial character of G). Let Go be the Galois group of some fixed algebraic closure of Q, and write'¢ : GQ -> Zp x for the homomorphism giving the action of GQ on µp"°, that is O'(() = (1/J(q) for all O' in GQ and all p-power roots of unity(. By Galois theory, G is a quotient of GQ, and the characters '¢k, where k is any even integer, factor through G. Then Iwasawa proved that there is a unique pseudo-measure Pp on G such that (k=2,4,6, ... );

(8)

this makes sense, since one can integrate any non-trivial p-adic homomorphism of G against any pseudo-measure. Let A(G)o = ker(A(G)-> Zp) be the augmentation ideal of A(G). Because Pv is a pseudo-measure, ppA(G)o is an ideal in A(G), and Iwasawa's "main conjecture" is the assertion that (9)

ppA(G)o

=Ji ... J,.,,A(G).

Familiarity should not allow us to forget how revolutionary (9) is. It gives a bridge between the p-adic analytic world, represented by the ideal on the left of (9), with the p-adic arithmetic world represented by the ideal on the right of (9). Outside the p-adic world, no one has even come close at present to formulating an analogue of (9) for the Riemann zeta function ( (s), nor of defining a good analogue of the right hand side of (9) for the maximal unramified abelian extension of the field generated over Q by all roots of unity. The honour of giving the first unconditional proof of (9) fell in 1984 to B. Mazur and A. Wiles [21], using methods from the theory of modular curves. But there is still great interest, especially for applications of similar ideas to elliptic curves, in studying the methods which led Iwasawa to (9), especially those in his wonderful paper {41 }. In this paper, Iwasawa used the classical cyclotomic units to construct a compact A(G)-module Yoo, together with a natural A(G)homomorphism

(10)

1{):

Yoo-> Xoo,

He then proved, by an ingenious use of an explicit reciprocity law going back to E. Artin and H. Hasse, that we have an isomorphism of A(G).modules

(11)

Y00

::;

A(G)/ppA(G)o;

here we are implicitly using the construction of Pp given in the later paper {43}. The precise definition of the module Y00 is as follows. Let Fn = Q((n + (; 1 ), where (n denotes a primitive pn+ 1-th root of unity (n = 0, 1, ... ). Let Vn be the subgroup of F,; generated by all conjugates of (1 - (n)(l - (; 1 ). The group of cyclotomic units Cn of Fn is defined to be the intersection of Vn with the unit

Iwasawa's work in algebraic number theory

:czi

group of the ring of integers of Fn. There is a unique prime Vn of Fn above p, and we write Un for the group of local units in the completion of Fn at Vn, which are congruent to 1 modulo Vn. Let Cn be the completion of Cn n Un in the Vn·adic topology. Then lwasawa's module Y 00 is the projective limit of Un/Cn, taken with respect to the norm maps. The unconditional proof of (11) was a great achievement. The beauty of (11) is that, via (8), it makes the values of the Riemann zeta function at the odd negative integers appear naturally in the arithmetic of cyclotomic fields. But (11) also leads inexorably to the formulation of the "main conjecture" (9). Indeed, assuming that the class number of the maximal real subfield of Q(µp) is prime top, lwasawa showed that the map cp appearing in (10) is an isomorphism, and hence (11) implies {9) in this case. It is plausible, but unknown at present, that cp is injective and has finite cokernel for all odd primes p (this turns out to be equivalent to Greenberg's conjecture that A = 0 for the cyclotomic Zp-extension F00 of the maximal real subfield of Q(µp)). Happily, the "main conjecture" is weaker, and boils down to proving that the kernel and cokernel of cp have roughly the same size (to be precise, have the same characteristic ideals) as A( G)-modules. In 1990, K. Rubin [24], using ideas on Euler systems inspired by the the work of F. Thaine [32] and V. Kolyvagin [16}, succeeded in proving just enough about the kernel and cokernel of cp to be able to deduce the "main conjecture" from (11). Another theme of Iwasawa's research, growing out of his work on (11), was his proof in {46} of a remarkable explicit formula for the Hilbert norm residue symbol in the field generated over Qp by the pn -th roots of unity for some integer n ;;Ji 1. His paper has inspired a large body of work .on explicit reciprocity laws in general over the last twenty years, beginning with the generalization of it to the fields generated by points of finite order on Lubin-Tate formal groups by A,. Wiles [33].

Influence of Iwasawa's work The heartland of arithmetic geometry is the study of the arithmetic of algebraic varieties defined over finite extensions of Q, arid the motives derived from them. Many of these arithmetic problems, at least conjecturally, are related to the special values of the L-functions attached to the cohomology of these varieties. The methods introduced by Iwasawa form the only general bridg~ known today between Ga.lois cohomological problems and the special values of £-functions. As such, they have already deeply influenced the course of arithmetic geometry over the last forty years, and it is highly likely that they will continue to do so in the years ahead, especially in the search for a proof of exact formulae, typified by the conjecture of Birch and Swinnerton-Dyer on the arithmetic of elliptic curves. For lack of space, we only briefly mention in what follows several areas in which lwasawa's ideas have already been fruitfully applied. We begin with the arithmetic of elliptic· curves. The first step in applying to these problems ideas· from lwasawa theory was made by B. Mazur [19]. Again, let .F be a finite extension of Q, and let E be an elliptic curve defined over F. Let Koo denote a Galois extension of F whose Galois group G is a p-adic Lie group of positive dimension (for example, K 00 could be the field obtained by adjoining to F all p-power ·roots of unity, or the coordinates of all p-power division points on

mi

John Coates

E). As was first emphasized by Mazur [19], the natural Iwasawa module to study in this situation is the analogue for K 00 of the classical Selmer group of E, which we denote by S(E/K00 ). By definition, S(E/K00 ) consists of all elements of H 1 (G(Q/K 00 ), Epoc) which, for each finite prime v of K 00 , localize to zero in the H 1 of the absolute Galois group of Koo,v acting on E(Koo,v); here Epoo denotes the Galois module of all p-power division points on E, and, as usual for infinite extensions, K 00 ,v is the union of the completions at v of all finite extensions of F contained in K 00 ; The natural action of G on H 1 (G(Q/K00 ),Epoo) induces an action of G on S(E / K 00 ). As S(E / K 00 ) is discrete and p-primary, it is more convenient to work with its compact dual

endowed with its acti(!n of the Iwasawa algebra A(G). It is always true that X(E/K00 ) is a finitely generated A(G)-module, but it is usually a difficult ques.: tion to determine 1ts rank as a A(G)-module. Mazur [19] raised the following two general questions. If pis any prime, and K 00 = F(µ,poo), prove that the group E(K00 ) of K 00 -rational points of E is always finitely generated as an abelian group? Secondly, if p is a prime such that E has good ordinary reduction at all primes v of F dividing p, and K 00 is the cyclotomic Zp-extension of F, prove that X(E/K00 ) has rank zero as a moduie over the Iwasawa algebra A(G)? We now know that both of these results are true when F = Q thanks to the deep work of K. Kato [15] and the fact that all elliptic curves o~r Q are now known to be modular,. but little has been prove:n about them over an arbitrary base field F. Kato's work is based on analytic methods in Iwasawa theory, and the first step towards creating such an. analytic Iwasawa theory for elliptic curves over Q was made by B. Mazur and H. Swinnerton-Dyer [20]. When Eis an elliptic curve defined over Q, and p is a prime where E has good ordinary reduction, they constructed in A(G), where K 00 is the cyclotomic Zp-extension of Q, a p-adic analogue of the complex £-function of E, and posed an analogue of the "main conjecture" (9) for X(E/K00 ). It is remarkable that up until now, all major progress towards proving "main conjectures" for the Iwasawa theory of elliptic curves uses· the same broad frame.. work of ideas as those introduced by Iwasawa in his work on cyclotomic fields, complemented· by additional· arguments using the notion of. Euler systems introduced by V. Kolyvagin [16]. Suppose first that F is an imaginary quadratic field with class number one, and E is an elliptic curve defined over F with complex mul~ ti plication by the ring O of integers of F (e.g. F = Q(v'=f), and E : ·y 2 = x 3 -Dx for D #- 0 in F). Let p be a prime ;;:i: 5, which splits in K, say p = 11'1i', and assume that E has good reduction at both 1r and 7i'. Let E1r"" be the group of all 1r-power division points on E, and take K 00 = K(E1roo ). Now S(E / K 00 ) is a p-prim.ary 0-module, and so it breaks up as the direct sum of its 71'-primary and it-primary subgroups. In fact, the 11'-primary subgroup of S(E/K00 ) is equal to Hom(G(M00 /K00 ),E1r00), where M 00 is the maximal abelian p-extensiori. of K 00 unramified outside 71'. In [2], [3], Andrew Wiles and I showed that one could carry through a modified form of the arguments of {41} for the A(G)-module G(M00 / K 00 ), provided one replaced the cyclotomic units by the elliptic units in the definition of Y00 • We proved the de$.ired analogue of (ll), and later K. Rubin [25] used Euler system arguments to deduce the analogue of the "main conjec-

Iwasawa's work in algebraic number theory

x:-ciii

ture" (9) in this case. In his thesis, R. Yager [36] extended Wiles and my method to the larger field K 00 = F(Ep°" ), proving a version of (11), and again K. Rubin [26] showed that the evident analogue of the "main conjecture" also holds in this case by rather more elaborate arguments with Euler systems. The motivation for proving these "main conjectures" for elliptic curves with complex multiplication is that even weak forms of them imply special cases of the conjecture of Birch and Swinnerton-Dyer (see [2], [3]). It was always known that it was a far more difficult problem to apply the ideas of {41} to elliptic curves E over Q without complex multiplication, largely because there was no longer a natural Euler system attached to E, which could also be proven to be related to the values at s = 1 of the twists of the complex L-functions of Eby Dirichlet characters. However, in a tour de force comparable in both ideas and technical brilliance to Iwasawa's original work, K. Kato [15] has recently proven the existence of such an Euler system for all elliptic curves E over Q. He has used this Euler system to prove that E(Q(µp°" )) is a finitely generated abelian group for all primes p (the proof uses also a result ofD. Rohrlich [22]). Suppose that pis a prime~ 5, where E has good ordinary reduction, and let K 00 be the cyclotomic Zp-extension of Q. Kato also proves that, in this case, X (E / K 00 ) is A( G)-torsion, and that a weak form of the Mazur-Swinnerton-Dyer analogue of the "main conjecture" (9) holds, in the sense that the ideal on the left of (9) is contained in the ideal on the right. We have only touched here on few of the many results which have now been proven about the Iwasawa theory of elliptic curves (see R. Greenberg [11] for more details). As an example of a different type of result from those mentioned above, we mention the remarkable paper by R. Greenberg and G. Stevens [13], which determines the first derivatives at the trivial character of the p-adic L-function of an elliptic curve E over Q, which has split multiplicative reduction at p, in terms of the Tate p-adic period qE attached to E over Qp. A quite different direction in which Iwasawa's work has been extended is the proof by A. Wiles [34] of the analogue of the "main conjecture" (9), when Q is replaced by an arbitrary totally real number field as base field. Here the proof of the existence of a good generalization of the pseudo-measure pp for totally real fields F was made by P. Cassou-Nogues [I], and P. Deligne and K. Ribet [7], after earlier slightly weaker results by J-P. Serre [29]. See also the remarkable paper by P. Colmez [5], which proves, for any totally real F, that this pseudo-measure has the residue one would expect at the trivial character, provided Leopoldt's p-adic regulator for F is non-zero. It seems highly unlikely that there exist natural Euler systems attached to totally real number fields, and in fact Wiles [34] has followed a quite different path from that of Iwasawa to prove the "main conjecture" in this case. His ingenious and deep method seems to have been inspired by the earlier work of K. Ribet [23], and exploits ideas of H. Hida [14] on p-adic families of modular forms. We end by indicating two other areas in which Iwasawa's ideas have played an important role. Firstly, nearly all that is known about the exact formulae for the orders of the higher even K-groups of the ring of integers of number fields in terms of zeta values, which have been conjectured by B. Birch and J. Tate [31], and S. Lichtenbaum [18], ultimately depends on some version of Iwasawa's "main conjecture". Secondly, in the work of M. Flach [9] and A. Wiles [35] on the exact order of the Selmer group of the symmetric square motive attached to a modular

uiv John Coates form, which plays such a key role in the proof that all elliptic curves over Q are modular, ideas from Iwasawa theory are a fertile ingredient.

References [l] P. Cassou-Nogues, "Valeurs aux entiers negatifs des fonctions zeta p-adiques", Inv. Math. 51 (1979), 29-59. [2] J. Coates, "Elliptic curves with complex multiplication and Iwasawa theory", Bull. London Math. Soc. 23 (1991), 321-350. [3] J. Coates, A. Wiles, "On the conjecture of Birch and Swinnerton-Dyer", Inv. Math. 39 (1977), 223-251. [4] J. Coates, A. Wiles, "On the p-adic £-functions and elliptic units", J. Australian Math. Soc. 26 (1978), 1-25. [5] P. Colmez, "Residu en s 371-389.

= 1 des fonctions zeta p-adiques", Inv.

Math. 91 {1988),

[6J A. Cuoco, P. Monsky, "Class numbers in Z/-extensions", Math. Ann. 225 (1981), 235-258.

[7] P. Deligne, K. Ribet, "Values at abelian £-functions at negative integers over totally real fields", Inv. Math. 59 (1980), 227-286. [8] B. Ferrero, L. Washington, "The Iwasawa invariant fields", Ann. Math. 109 (1979), 377-395.

/J,p

vanishes for abelian number

[9] M. Flach, "A finiteness theorem for the symmetric square of an elliptic curve", Inv. Math. 109 (1992), 307-327. [10] R. Greenberg, "On the Iwasawa invariants of totally real number fields", Amer. J. Math: 98 (1976), 263-284. [11] R. Greenberg, "lwasawa theory for elliptic curves", in Springer Lecture Notes 1716, 51-105. [12] R. Greenberg, "Iwasawa theory-past and present", in Adv. Study in Pure Math., Vol. 30, Class Field Theory - its centenary and prospect, Math. Soc. Japan (2001), 335-385. [13] R. Greenberg, G. Stevens, "p-adic £-functions and p-adic periods of modular forms", Inv. Math. (1993), 407-447. [14] H. Hida, "Iwasawa modules attached to congruences of cusp forms", Ann. Sci. Ee. Norm. Sup. 19 (1986), 231-273. [15] K. Kato, "p-adic Hodge theory and values of zeta functions of modular forms", to

appear. [16] V. Kolyvagin, "Euler systems", in Progress in Mathematics 87, Birkhauser (1990), 435-483. [17] T. Kubota, H. Leopoldt, "Eine p-adische theorie der zetawerte", Orelle 214 (1964), 328-339.

[18] S. Lichtenbaum, "Values of zeta functions, etale cohomology, and algebraic Ktheory", in Springer Lecture Notes 342, 489-501.

Iwasawa's work in algebraic number theory

XXV

[19] B. Mazur, "Rational points of abelian varieties with values in a tower of number fields", Inv. Math. 18 (1972}, 173-266. [20] B. Mazur, H. Swinnerton-Dyer, "Arithmetic of Weil curves", Inv. Math. 25 (1974), 1-61. [21] B. Mazur, A. Wiles, "Class fields of abelian extensions of Q', Inv. Math. 76 (1984), 179-330. [22] Rohrlich, D., "On £-functions of elliptic curves and cyclotomic towers", Inv. Math. 75 (1984), 409-423. [23] Ribet, K., "A modular construction of unramified p-extensions of Q(µp)", Inv. Math. 34 (1976), 151-162. [24] K. Rubin, "The main conjecture", Appendix to "Cyclotomic Fields I-II" by S. Lang, Graduate Texts Math. 121, Springer (1990}. [25] K. Rubin, "The one variable main conjecture for elliptic curves with complex multiplication", in London Math. Soc. Lecture Notes 153, 353-371. [26] "K. Rubin, "The "main conjecture" of Iwasawa theory for imaginary quadratic fields", Inv. Math. 103 {1991), 25~68. [27] J-P. Serre, unpublished letters to K. Iwasawa dated 12th and 19th August, 1958. [28] J-P. Serre, "Classes des corps cyclotomiques" (d'a.pres K. Iwasawa), Sem. Bourbaki 174 (1959). [29] J-P. Serre, "Formes modulaires et fonctions zeta p-adiques", Springer Lecture Notes 350, 191-268. [30] W. Sinnott, "On the µ-invariant of the r-transform of a rational function", Inv. Math. 75 (1984), 273-282. [31] J. Tate, "Symbols in arithmetic", Actes, Oongres Intern. Math., 1970, tome 1, 201-211. [32] F. Thaine, "On the ideal class groups of real abelian number fields", Ann. Math. 128 (1988), 1-18. [33] A. Wiles, ''Higher explicit reciprocity laws", Ann. Math. 107 {1978), 235-254. [34] A. Wiles, "The Iwasawa conjecture for totally real fields", Ann. Math. 131 {1990}, 493-540. [35] A. Wiles, "Elliptic. curves ahd Fermat's Last Theorem.", Ann. Math. 141 (1995), 443-551. [36] R. Yager, "On two variable p-adic £-functions", Ann. Math. 115 (1982), 411-449.

[1] Ueber die Struktur der endlichen Gruppen, deren echte Untergruppen sämtlich nilpotent sind Proc. Phys.-Math. Soc. Japan, 23-1 (1941), 1-4.

(Gelesen am 21. Dezember, 1940.)

Eine Gruppe @ heisst nilpotent, w~1111 die aufsteigende Zentralreihe von @ die ganze Grr!-ppe als letztes Glied enthält, wenn also · l=äoC 31C ..... C 3c=@

ist, wobei 3,~1/31 immer das Zentrum von @{a, ist• .c bedeutet natürlich eine ganze rationale Zahl. Eine endliche nilpotente Gruppe ist bekanntlich das direkte Produkt ihrer Sylowgruppen und umgekehrt'2'. Im folgenden soll die Struktur derjenigen endlichen Gruppen untersucht werden, deren echte Untergruppen alle nilpotent sind. 1. @ s!.'li also eine endliche Gruppe, deren echte Untergruppen alle nilpotent sind, und p sei ein Primteiler der Ordnung g von @. Wir nehmen zunächt an, dass @ nicht p-normal ist, d.h. dass das Zentrum 3„ einer p-Sylowgruppe ~ von @ als Nicht-Normalteiler in einer anderen p-Sylowgruppe ~ enthalten ist)=@ sein. @ enthält also einen Normalteiler ~ mit einer p-Potenz Ordnung. Nunmehr sei @ p-normal, und ai> nicht ein Normalteiler von @. Der Norm.alisator N(a,,) von 311 ist dann als echte Untergruppe von @ nilpotent., und seine p-Faktorkommutatorgruppe< 6> ist nicht die Ein· b.eitsgruppe. Nun gilt n&eh Grünrs, die folgende Tatsache: Wenn die endliche Gruppe @ p-normal ist, dann ist die p-Faktorkommutatorgruppe von ~ isomorph zu der p-Faktorkommutatorgruppe des. Normalisators eines p-Zentrums SP· Nach diesem Satz enthält @ also (1) H. Zassenhaus, Lehrbuch der Gruppent4eorie I, (1937), S. 105. Zitiert mit H.Z.. (2) H.Z., S. 107.

(3) 0. Grün, J. reine u. angew. Math. 174, S. 1-14, od. H.Z., S. 135. (4) W. Burnside, Theory of G1·011ps of :finite order, 2nd ed. (1911), S. 1/S6, (5) H.Z., S. 122.

(6) Vgl, (3).

1

[1]

2 2

Kl•ukiti IW ASA\V .A,

[Vol. 23

-eineu Normalteiler vou einem p-Poteuz Index-. In jedem Fall enthält @ daher einen Normalteiler entweder von einer p-Potenz Ordnung oder von einem p-Potonz Index, Da auch jede Faktorgruppe von @ die Eigenschaft besitzt, dass ihre sämtlichen echten Untergruppen nilpotent sind, so ergibt sich hieraus, dass @ auflösbar ist. Also gilt Satz 1. Jede endliche Gruppe @, deren echte Untergruppen sämtlich nilpotent sind, ist auflösbar. Nun sei g=p?p~2 •• •• p~r mit rie!:3, wobei p, verschiedene Primteiler vou g sind. In der endlichen auflösbaren Gruppe @ gibt es nach P. Han, Das Zentrum von SL(n, K) besteht aus allen lE mit J... =1, wenn mit E die identische Transformation · bezeichnet wird. Die Faktorgruppe von SL(n, K) nach dem Zentrum heisst die spezielle projektive Gruppe PSL(n, K). Für n > 1 ·ist nun PSL(n, K) eine einfache Gruppe, ausser in zwei Fällen n=2, K=GF(2) und n.=2, K=GF(3). Dieser Satz wurde zuerst von L. E. Dickson für endliche Grundkörper K bewiesen8>, Sein Beweis aber wird, wie van der Waerden bemerkt4>, durch eine kleine Abänderung auch für den Fall unendlicher Grundkörper K gültig, wenn K von der ·Charakteristik ¾: 2 oder vollkommen ist. In der vorliegenden Note soll dieser Satz ohne oben erwähnte Einschränkung über K neu bewiesen werden. Hilfssatz 1. SL(n, K) fällt mit seiner Kommutatorgruppe zusämmen, ausser wenn n=2, K=GF(2) oder n=2, K=GF(3). Beweis. Zunächst sei n > 3 und i, j (1 ;;;;;; i, j< n) beliebig gegeben. Für k ¾: i, j gilt dann B,, k. ABk. i, 1Bi~1,.. 1Bi,1;, 1=B„ j, A , Damit ist unsere Behauptung erledigt. 1) Bzgl. der Bezeichnungen GL(n, K), SL(n, K) u.s.w. vgl. B. L. van der Waerden: Gruppen von linearen Transformationen, 1935. . 2) . Vgl. L. E. Dfokson: Linear groups, witb an exposition of Galois Field theoey, 1901, S. 78. . 3) VgL loc. cit. 2), S. 88. 4) Vgl. loc. cit. 1), S. 7,

5

6

[2] K.

58

Nun sei n=2. ('A.

0

(Vol. 17,

!WASAWA.

Alsdann ist

0 ) (1 µ) (J- 0) (1 -p)=(l µ(i. -1)). 1

0 1

).-l,

0

2

O

J

1

0

1

·wenn K nicht GF(2) oder GF(3) ist, so gibt es it1 K mindestens ein Element Ä =½= 0 mit ;.2 ~ 1. Da p ein beliebiges Element aus K ist, so enthält die Kommutatorgruppe von SL(2, K) alle B 1, 2, ,1 und ebenso alle B2.1,;.: sie fällt also mit SL(2, K) selbst zusammen. Die Gruppe PSL(n, K) besteht bekanntlich aus projektiven Transformationen des (n-1)-dimensionalen projektiven Raumes über K. Wir können also diese Gruppe als Permutationsgruppe von Punkten in P..,.i(K) auffassen. Es gilt dann Hilfssatz 2. PSL(n. K) ist (n > 1), als Permutationsgruppe, zweifach transitiv. Beweis. Es seien a, b zwei beliebige Punkte von P.,._ 1(K): a=(a1,%··•,o.,,,),

b=(ß1,ß2, ... ,ß,.).

Es .genügt ersichtlich nur zu zeigen, dass es eine solche Transformation in PSL(n, K) gibt, die Punkte e1 =(1, {), 0, ... , 0), ¾=(O, l, 0, ... , 0) in a bzw. b übeiführt. Wegen a =½= b gibt es dann eine geeignete Matrix

(7

Ao=

(;; ~:·:::::) : :

a„ ßn ••...

mit ! A0 1~ 0. r1 wird da.nn wie ersichtlich durch die Matrix A0 /I A 0 \ angegeben. Nunmehr sei ,)) diejenige Untergruppe von @=PSL(n, K) (n > 1), die aus allen Transformationen aus @ besteht, die e1 festlassen. Eine Transformation aus ~ ist daher von der Form

-( r

A-::

°-2"'' ..

A1

·a,.)

0

wobei A1 eine (n-1)-reihige Matrix mit alle MatrizenD der Form

ad A1 1=1

ist.

In

~1

bilden

einen Normalteiler 910, denn genau diese Matrizen werden beim Homomorphismus A---> A1 der (n-1)-reihigen Einheitsmatrix E,i.-1 .zugeordnet. Bi, 2,;. sind sämtlich in 910 enthalten. --------------+----------···---1) Ausführlicher:

die durch diese

SL(n, K)/8,.,;, PSL(n, K).

Matrizen vertretenen

Restklassen

von

[2]

7 No. 3.]

Über die Einfachheit der speziellen projektiven Groppen.

59

Nun sei m ein Normalteiler von @, der keine Einheitsgruppe ist. Da @ nach Hilfssatz 2 zweifach transitiv ist, so ist m als Permutationsgruppe von Punkten aus P,,._ 1(K) sicherlich transitiv. Es ist also @=9l.l), und 9l* =~Rmo ist natürlich ein Normalteiler von @. Da ein beliebiges Bi, 1, ; mit B1. 2. ±A in SL(n, K) konjugiert ist, so enthält 9l* mit B1, 2, ; zugleich alle Bi, :i, ; ; in* fällt also mit @ zusammen. Aus @= 9l~0 folgt @/m ~'ll0 /W r. ~o und '1l ist mithin ein Normalteiler mit abelscher Faktorgruppe. Da aber @ nach Hilfssatz 1 ausser in zwei genannten Fällen mit seiner Kommutatorgruppe zusammenfällt, so ergibt sich hieraus @ ='ll. Es gilt also der Hauptsatz. Die Gruppe PSL(n, K) ist einfach (n > 1), ausser wenn n=2, K=GF(2) oder n=2i K=GF(3). Für n=2, K=GF{2) bzw. n=2, K=GF(3) ist PSL(n, K) bekanntlich isomorph mit der symmetrischen Gruppe von 3 Symbolen bzw. mit der Tetraedergruppe. In ähnlicher Weise kann man auch die Einfachheit der projektiven Komplexgruppen PC(2m, K) 1> nachweisen. 1) PC(2m, K) ist als Permutationsgruppe von Punkten aus P2m-1{K) nicht zweifach transitiv. Es .Jässt .sich aber leicht beweisen, dass. jeder Normalteiler von PC(2m, K) transitiv ist, was für den Beweis der Einfachheit wesentlich ist.

[3]

••

Ober die endlichen Gruppen und die Verbände ihrer Untergruppen

J. Fac. Sei. Tokyo lmp. Univ., Sect I. 4-3 {1941), 171-199.

Die Untergruppen einer gegebenen Gruppe @ bilden bekanntlich einen Verband ; bezeichnet man nämlich für zwei beliebige Untergruppen m, ~ von @ die von m: und ~ erzeugte Untergruppe bzw. den Durchschnitt von ~, ~ mit ~ v ~ bzw. W bezeichnet, so sieht. man leicht ein, dass alle Verbandsaxiome erfüllt sind. Wir bezeichnen diesen Verband mit V(@) und nennen ihn den zu @ gehörigen Verband; Nun wollen wir folgende Definition einführen: Definition. Eine Gruppe @ heisst modular, distributiv, komplementär bzw. Boolesch, wenn der zugehörige Verband V(@) modular, distributiv, komplementär bzw. Boolesch ist. Wir bezeichnen oft solche Gruppen bzw. auch kurz mit M-, D-, K bzw. B-Gruppe. Wenn ferner V(@) modular komplementär und irreduzibel ist, also die „ projektive Geometrie" bildet, dann sagen wir, @ sei eine P-Gruppe. Im folgenden soll die Relation zwischen @ und V(@) untersucht werden; insbesondere soll die Struktur der endlichen M-Gruppen (§§ 1 u: 4), D-, B- bzw. P-Gruppen (§ 2) vollständig bestimmt werden. Daran anschliessend wird auch die Struktur der endlichen „ quasiHamiltons.chen" Gruppen bestimmt (§ 3), die eine Verallgemeinerung der Hamiltonschen Gruppen bilden.

n~

§ 1.

1. Ein Verband A heisst ausgeglichen, wenn alle Hauptkette, die zwei Elemente von A verbinden, gleiche Länge haben.° Da ein modularer Verband stets ausgeglichen ist, so untersuchen wir zuerst die Struktur derjenigen endlichen Gruppe, deren zugehöriger Verband ausgeglichen ist. Hilfssatz 1.2> Wenn der zugehörige Verband V(@) einer endlichen Gruppe @ ausgeglichen ist, ~ hat @ eine Hauptreihe (1)

1 = @o C@1 C · · · C@,. =@ ,

1) G. Birkhofl', Proc. Camb. Phil. Soc. 29, 441-464, 1933. 2) Vgl. 0. Ore, Duke Math. Jour., 5, 431-460, 1939.

8

(3]

9 Kenkiti Iwasawa.

172

wobei [@i :@i-1]=pi (i= 1, 2, ... , r) Primzahlen mit P1> P2> ···>Pr sind. Beweis. Für eine Gruppe von einer Primzahlordnung ist die Behauptung klar. Wir beweisen also den Hilfssatz durch vollständige Induktion nach der Ordnung von @. g sei also die Ordnung von @ und Pr der kleinste Primteiler von g. Wir nehmen zunächst an, dass @ Pr-normal3> ist. Nach Grün4> ist dann die p,.-FaktorkommutatorNz/N;(pr) von gruppe @j@1(pr) von @ zur Pr-Faktorkommutatorgruppe I Nz isomorph, wobei .8 das Zentrum einer Pr-Sylowgruppe ~ von @, und Nz der Normalisator von .8 bedeutet. Ist .8 nicht ein Normalteiler von @, so ist Nz eine echte Untergruppe von @, und nach Induktionsvoraussetzung ist @j@'(pr) '::::'..N,/Nipr) offenbar keine Einheitsgruppe, Sei nun @ nicht Pr-normal. Nach einem Satz von Burnside5>. gibt es dann eine Pr-Untergruppe ~ von @, so dass der Normalisator N(~) von ~ ein Element A mit ~iner zu Pr primen Ordnung enthält, das in .p einen nicht identischen Automorphismus a ; (2)

H"=AHA- 1 ,

He~,

induziert. Wäre nun N(Sj) eine echte Untergruppe von @, so folgte aus der Induktionsvoraussetzung, dass die Untergruppe ~ 1 = {A, ~} direktes Produkt von .p und {A} ist. Awäre dann alsö mit~ elementweise vertauschbar, was mit (2) in Widerspruch steht. ~ ist daher ein Normalteiler von @. In jedem Fall enthält @ also einen Normalteiler 91 von einer Primzahlpotenzordnung oder von einem Primzahlpotenzindex. Da offenbar mit V(@) auch V(\.lc) und V(@/\.lc) ausgeglichen sind, so sieht man nach Induktionsvoraussetzung sofort, dass @ auflösbar ist. Da @ auflösbar ist, so ist eine Kompositionsreihe von@ offenbar zugl~ich eine Hauptkette von V(@), die @ mit 1 verbindet, und die Länge r dieser Hauptkette ist gleich der Anzahl der Primfaktoren von @. Da jede Hauptkette, die @ mit 1 verbindet, stets· gleiche Länge r hat, so ergibt sich hieraus, dass für zwei Untergruppen i!(, ~ der Index.[i!C: 5B] eine Primzahl ist, wenn sie in einer Hauptkette von V(@) benachbart sind, d. h. wenn 5B eine maximale Untergruppe von i!C ist. Es sei nun p1 der grösste Primteiler von g und ~ 1 sei eine p1-Sylowgruppe von @. Wir verfeinern die Kette 1 c::: ~1 C @ ZU einer Hauptkette . 1 =mo C::: m1 C ··· C:: mk=~1 C::: mk+l C::: ••• C5!3r=@. [mk+i: ~1]=q ist dabei, wie schon bemerkt, eine Primzahl. 3) 0. Grün, Jour. f. reine u. angew. Math. 174, 1-14, 1935. 4) Vgl. Grün, a. a. Q. 3). 5) W. Burnside, Theory of groups of finite Order (1911), S. 156.

Die

10

[3] Über die endlichen Gruppen und die Verbände ihrer Untergruppen.

173

Anzahl s der Pi-Sylowgruppen von mk+1 ist also entweder 1 oder q. Da aber nach Annahme Pi.>q, und nach dem Sylowsatz S=l mod. p ist, so muss s= 1 sein. ~1 ist also als Normalteiler in Q.h+1 enthalten. [mk+2 : mk+1] = q' ist nun auch eine Primzahl mit Pt> q', also ergibt sich in genau derselben Weise wie oben, dass ~ 1 als Normalteiler in mk+2 enthalten ist. So fortfahrend beweist man schliesslich, dass ~1 ein Normalteiler von @ ist. Sei nun ,ßi das Zentrum von ?l31, und ~1 eine Untergruppe von @ von der Ordnung [@ : ?l31]. Dass es wirklich ein solches .P1 gibt, folgt daraus dass @ auflösbar ist. 0> Verfeinert man wieder die Kette 1 C~1 C 31~1 C@ zu einer Hauptkette

l=mocm1 C ··· C ma=~1 cma+1 C ····C$z=ßi-P.1 C ··~ cmr=@, so ist der Index [m/i+l: ~1] offenbar gleich p. 911 =mk+l (""\ .81 ist also

ein Normalteiler von mk+ 1 mit der Ordnung p1, und da es in 31 enthalten ist, so ist es auch. ein Normalteiler von ~1, folglich von @. Wendet man nunmehr die Induktionsvoraussetzung. auf @/911 an, so sieht man leicht ein, dass die Behauptung für @ richtig ist. Hüfssatz 2.1> Dann und nur dann ist der Verband V(@) einer endlichen Gruppe @ ausgeglichen, wenn für zwei Untergruppen 2{, ~ von @ der Ind,ex [m : ~] stets eine Primzahl ist, sobald ~ als maximale Untergruppe in menthalten ist. Beweis. Dass die Bedingung notwendig ist, haben wir schon im Beweis von Hilfssatz 1 gesehen. Ist umgekehrt die Bedingung erfüllt, so ergibt sich sofort, dass die Länge einer Hauptkette, die zwei Untergruppen 21, ~ verbindet, immer der Anzahl .der Primfaktoren vom Index [~: ~] gleich ist. V(@) ist also ausgeglichen. Aus Hilfssätzen 1, 2 folgt nun Satz 1. Es sei @ eine endliche Gruppe und V(@) der zugehörige Verband. V(@) ist dann und nur dann ausgeglichen, wenn @ eine primzahlstufige Hauptreihe besitzt, wenn es also eine solche Hauptreihe (3)

gibt, wobei die Indizen [91.: WH] (i=l, 2, ... , r) sämtlich Primzahlen sind. Beweis. Dass @ eine solche Hauptreihe besitzt, falls V(@) ausgeglichen ist, folgt aus Hilfssatz 1. Nach Hilfssatz 2 haben wir also nur noch zu beweisen, dass für zwei Untergruppen m, $ einer Gruppe @ mit der Hauptreihe (3), der Index [~ : ~] stets eine Primzahl ist, 6) P. Hall, J. of London Math. Soc. 12, 198-200, 1937. 7) Vgl. 0. Ore, a. a. 0. 2).

11

[3] 174

Kenkiti lwasawa.

wenn 5B als maximale Untergruppe in menthalten ist Dazu wenden wir Induktionsschluss nach der Ordnung von @ an. -Da mit @ auch jede ·Untergruppe von @ eine primzahlstufi.ge Hauptreihe besitzt, so genügt es nach Induktionsvoraussetzung nur zu ~eigen, dass eine maximale Untergruppe ~ von @ einen Primzahliridex hat. Wenn ~ 9l1 enthält, so wenden wir Induktionsvoraussetzung auf @/fil1 ap. Wenn aber llR 9l1 nicht enthält, so ist im911=@, l!Rn911=l. Der Index[@: IDl] ist dann gleich der Ordnung von 911·, also eine Primzahl, w. z. b. w.. Aus diesem Satz folgt nun, dass eine Gruppe @ mit einer primzahlstufigen Hauptreihe sogar eine solche Hauptreihe besitzt, wie sie im Hilfssatz 1 angegeben ist. Das ist aber natürlich auch direkt leicht zu beweisen. 2. Wir untersuchen nun die Struktur· der endlichen M-Gruppen und zeigen, dass eine beliebige endliche M-Grüppe ins direkte Produkt ebensolcher Gruppen zerlegbar ist, deren· Ordnungen ·durch höchstens zwei verschiedene Primzahlen teilbar sind . . In dieser. Nr. bedeutet @ immer eine endUche M-Gruppe. Die Primfaktoren der Ordnung von @ bezeichnen wir mit p, q, ... Hilfssatz 3. Eine endliche M-Gruppe ist auflösbar. Beweis. Folgt sofort aus Satz 1. · · Hilfssatz 4, Wenn die Ordnungen, a, b von zwei Untergruppen m, 5B von ·@ teilerfremd sind, so ist die Ordnung ·von mu 5B gleich ab. Beweis. Die Ordnung von %t v 5B ist offenbar durch ab teilbar. Da für einen . modularen Verband V(@) die Verbandsisomorphie %t v 5B / ~ ~ ~ / %t n 5B gilt,S> so ist die Länge der Hauptkette, die %t v 5B mit %t verbindet; gleich der Länge der Hauptkette, d.ie 5B mit %tri 5B= 1 verbindet. Nach Hilfssatz 2 ist sodann die Anzahl der Primfaktoren der Ordnung von & v 5B gleich der Summe von der Anzahl der Primfaktoren der Ordnungen von ~ und der von 5B. Die Ordnung von 21 v 5B ist also gleich ab. HilfBBatz 5. · Es sei ~ eine P·Untergruppe, und O eine q-Untergruppe von @. Wenn p grösser als q ist, dann ist O im Normalisator N(~) von ~ enthalten. Beweis. Wenn die Ordnung von ~ bzw. D p« bzw. qfl ist, so ist die Ordnung von ~ v O· nach Hilfssatz 4 gleich paqfl, Nach Hilfssatz 1 ist dann p-Sylowgruppe ~ ein Normalteiler von ~ u 0, also ist 0 in N(~) enthalten. Für modul.aren p-Gruppen sollen nun einige ejnfache Tatsachen 8) Vgl. G. Birkhoff, _Lattice theory (1940), S. 37.

[3]

12 Über die endlichen Gruppen und die Verbände ihrer Untergruppen.

175

hergeleitet werden, die wir bald nötig. haben. Eine ausführliche Untersuchung darüber soll in § 4 vorgenommen werden. Hilfssatz 6. Für zwei beliebige Untergruppen. fil, ~ einer modularen p-Gruppe JM gilt (fil) (~) =(fil "~) (2{ u ~}. Gleiches gilt für ~. u. s. w.. Beweis. Nach der Verbandsisomorphie ~v~/~(~~/filr't~ enthalten die lndizen [~{ v~: ~] und [~: ~ r't~] gleiche Anzahl von Primfaktoren. Da sie aber beide p-Potenzen sind, so. sind sie einander gleich. Hüfssatz 7. Zwei beliebige Elemente A, B von · der Ordnung p aus einer modularen p-Gruppe @ sind stets vertauschbar. Beweis. Die Gruppe {A,. B} ist stets abelsch, da ihre Ordnung ist. nach Hilfssatz 6 p oder Nun soll die Struktur der nicht nilpotenten M-Gruppe untersucht werden, deren Ordnung .· durch genau zwei verschiedene Primzahlen teilbar ist. p und q SElien vorläufig immer zwei veJ;"schiedene Primzahlen. Hilfssatz 8. · @. sei von der Ordnung paq. Falls @ nicht nilpotent ist und q > p gilt, dann ist jede p-Sylowg_ruppe von @ zyklisch. Beweis. Im Fall a=l ist die Behauptung klar. Sei a=2. Nach Hilfssatz 1 folgt aus q>p, dass eine q-Sylowgruppe ·O ein Normalteiler von @ ist. Nach Voraussetzung ist dann eine p-Sylowgruppe kein Normalteiler von @. ~ sei ein maximaler Durchschnitt zweier verschiedener p.:.Sylowgruppen: m=~1 r't ~2, Nach der Verbandsisomorphie ~1u~d\ß1~5ß2/mJst meine maximale Untergruppe von ~2, d.h. [~2:~)=p, da ?ß1 eine maximale Untergruppe von @=?lhu~a ist. ~ ist als Durchschnitt sämtlicher p-Sylowgruppen von @ ein Normalteiler,9' folglich mit O elementweise vertauschbar. Wir. nehmen nun an, dass eine p-Sylowgruppe ~ nicht. zyklisch ist. Setzt man (fil) bedeutet dabei die Ordnung von fil.

r

O={Q}, ~={D}, ~={D,P}, ©={P,Q}, so ist @=© xSi), Nach Voraussetzung ist © nicht abelsch, also ist die Ordnung von PQ gleich p. Wenn ~l={PQD}, ~={P} gesetzt wird, so ergibt sich aus der Verbandsisomorphie fil u~/ 2{~~. dass die Ordnung von. 2{ u~ gleich oder pq ist, was aber wegen

r

2( u

~ = {PQD,

P} = { QD, P} = {Q, D, P} =@

9) Vgl. z.B. H. Zassenhaus, Lehrbuch der. Gruppentheorie (1937), S. 102. (zitiert mitH.Z.).

[3] 176

13 Kenkiti Iwasawa.

ein Widerspruch ist. lß muss daher zyklisch sein. Sei nun a > 3; Wie im Fall a=2 ist der Durchschnitt 1:J aller p„Sylowgruppen ein Normalteiler von @. Es gilt wie vorher [~: 1:J]=p, und SD ist mit O elementweise vertauschbar. Wir nehmen wieder an, dass lß nicht zyklisch. ist, und wählen eine von SD verschiedene Untergruppe lß1 mit [lß :~1]=p. Setzt man nun SD1 =SD {"'\ lß1, so ist SD1 ein Normalteiler von @, und @/1:J1 ist eine M-Gruppe von der Ordnung p 2q. Da eine p-Sylowgruppe ~/~ von @/SD1 nicht zyklisch ist, so muss @/SD1, abelsch sein. ~ ist folglich ein Normalteiler von @, was aber unserer Voraussetzung widerspricht. ~ muss_ also wieder zyklisch sein. Na.eh diesem Hilfssatz ergeben sich nun die folgenden wichtigen Hilfssätze 9, 10. Hüfssatz 9. Wenn eine p-Sylowgruppe ~ kein Normalteiler von @ ist, so ist ~ zyklisch. Beweis. SD sei der Durchschnitt zweier verschiedenen p"'8ylowgruppen ~ 1, ':ß2 von @, und ij2 sei eine 1:J enthaltende Untergruppe von ':ß2 mit dem Index [~2.: 1:Jl =p. Setzt man nun ~ =~1 u 2, so folgt aus · SD=~2 {"'\ ~1 und ~/~1~~2/SD, dass der Index [~: ':ß1l;,,,q eine Primzahl ist. (Vgl. Hilfssätze 1, 2). ~ ist also eine Untergruppe von der Ordnung 'J)"q, und ?ß1 ist kein Normalteiler von :p. Nach Hilfssatz 8 ist ~1 zyklisch.

m

,,

..

.

Hilfssatz 10, ~ 1 und ?ß2 seien zwei beliebige p-Sylowgruppen von @. Wenn ~1 nicht mit ~ 2 zusammenfällt, so ist der Index von m=~1{"'\~2 in ':ß1 bzw~ in lß2 gleich p. Es gfft dann Q':ß1Q- 1=lß2 mit einem geeigneten .Element Q, dessen Ordnung eine .Primzahl q (:¾= p) ist. Beweis. ~2 sei eine SD enthaltende Untergruppe von ~2 mit dem Index [~2 : SD]=p. Wenn ~=~ 1 u~ 2 gesetzt wird, dann zeigt sich genau wie im Beweis yon. Hilfssatz 9, dass [~: ~J=q eine Primzahl ist. Der Durchschnitt ~ aller p-Sylowgruppen von ~ ist, wie beim Beweis von Hilfssatz 8 gezeigt wird, ein Normalteiler mit [~: ~]=p. Wenn ~2 nicht mit lß2 zusammenfällt, .s.o ist ij2 in SD enthalten, weil ~ 2 nach Hilfssatz 9 zyklisch ist. ij2 muss. dann auch in ~1 enthalten

14

~] Über die endlichen Gruppen und die Verbände ihrer Untergruppen.

177

sein, was mit SD=~2r\ ?ß1 in Widerspruch steht. Es folgt also t2=)ß2 und [~1: ~]=(?ß2: ~]=p. Da ?ß1t )ß2 zwei p-Sylowgruppen von © sind, so gibt es in © ein geeignetes Element Q von der Ordnung q, so dass Q?ß1Q- 1= )ßz ist. Nun soll die Struktur der nicht nilpotenten M-Gruppen von der Ordnung paqß bestimmt werden. Wenn p grösser als q ist, so ist eine p-Sylowgruppe ?ß ein Normalteiler von @. (Vgl. Hilfssatz 1). Nach Voraussetzung ist dann eine q-Sylowgruppe D kein Normalteiler, also nach Hilfssatz 9 zyklisch: D= {Q}. Der Normalisator von D in @ sei mit N(D) bezeichnet. Wir nehmen zunächst an, ' dass N(D) :¾= D ist. Die Untergruppe ~=N(D)" ?ß ist dann keine Einheitsgruppe und ist offenbar mit :0 elementweise vertauschbar. A sei nun ein Element aus ~ und B ein beliebiges Element aus )ß, das nicht in {A} enthalten ist. Wenn die Ordnungen von A und B beide gleich p sind, so ist auch die Ordnung von AB gleich p. (Vgl. Hilfssatz 7). Nach Hilfssatz 5 ist D im Normalisator von {B} bzw. {AB} enthalten. Ist nun also so folgt aus QAQ- 1 =A, dass r=a-l mod. p ist, d. h. dass B mit Q vertauschbar ist. Sei Ö eine von D verschiedene q-Sylowgruppe von @, so ist D=PDP- 1 mit einem Element P von der Ordnung p (Hilfssatz 10). Da andererseits jedes. Element B von der Ordnung p mit Q vertauschbar ist, so ergibt P:OP-L=fö einenWidersp:ruch. Der Normalisator N(D) muss also mit D zusammenfallen. Pi~ 1 sei nun ein beliebiges Element von ?ß. Da ADP11 =\= :0 ist, so gibt es nach Hilfssatz 10 ein Element P von der Ordnung p, so dass P 10f11 =PDP-1 gilt. P11P liegt daher in N(D)n~=l, also gilt P1 =P. Jedes Element. von ?ß ausser Eins muss mithin die Ordnung p haben. Nach Hilfssatz 7 ist ?ß daher eine abelsche Gruppe vom Typus (p, p, ... , p). Für zwei be~ liebige Elemente Ai. B 1 von jß beweist man genau in derselben Weise wie oben, dass aus QA 1Q-1 =Af, QB1Q-1 =Bf folgt P=P mod. p. Es gilt. also für ein beliebiges Element P von $ die Relation

wobei r eine bestimmte nur von @ abhängige natürliche Zahl ist. Da ?ß nicht mit D, wohl aber mit {Qq} elementweise vertauschbar ist, so muss dabei r ~ 1 mod. p, rq=l mod. p sein. Dass umgekehrt eine Gruppe @ von der Ordnung paqfl eine MGruppe ist, falls sie die oben genannte Struktur besitz~, kann man

15

[3] 178

Kenkiti Iwasawa.

fQlgendermassen beweisen. t(, Q3, [ seien beliebige Untergruppen von @ und 'l( sei in Q: enthalten. Es soll dann gezeigt werden, dass (2Cv~)0[=2Cv(Q3n[) gilt. Dazu .genügt es offenbar nur zu beweisen, dass (2C v Q3) f"'\ [ in 2{ v ()ß n [) enthalten ist, und kann man dabei. noch annehmen, dass 21 bzw. )ß kein Normalteiler von @ ist, in unserem Fall also eine q-Sylowgruppe von @ enthält. Man setze

2C=2Lp0, )ß=~pÜl,

[=[PD,

wobei ~(p, Q3ri bzw. G:v eine p-Sylowgruppe von 2(, ?S bzw. [ mit &v C [P ist und O und 0 1 q-Sylowgruppen von@ sind. Ferner sei 0 1 =PDP- 1 mit einem Element P aus ~. Jedes Element T aus (& v 58) f"'\ [ lässt sich dann in der Form

schrieben, wobei A, B bzw. C geeignetes Element ·von 2fm )SP bzw. [v ist. Sei zunächst D1 in G: enthalten und P ein Element aus [p, · Dam1 ist m" [ = (~p ("\ [p)Ü1, also & V ()ß f"'\ @:) = {2C P• (Q3p 0 [p), P} . 0, und T ist offenbar in 2C v ()Sn[) enthalten. Wenn aber 01 nicht. in [ enthalten ist, so ist 5B n@: =()ßp r"1 @:p) · { Qq}, folglich 2C v (5B n [) = {2Cp v ()ßp" ffp)}D. In diesem Fall liegt P nicht in [p v mp (sonst kann man D1 in [ wählen), also e-1A- 1C=P11 =l, und B=A- 1Ce [pn)ßv, T= ABQ" ist wieder in & v (~ n.[) enthalten. Somit gilt der Satz fJ. Eine nicht nilpotente Gruppe @. von der Ordnung paqB (p>q) ist dann. und nur· dann modular, wenn sie folgende Struktur besitzt: (1) Eine p-Sylowgruppe ~ ist ein Normalteiler von @. Sie ist eine abelsche Gruppe vom Typus (p 1 p, ... , p ). (2) Eine q-Sylowgruppe D ist zyklisch; 0 = {Q}. (3) Für ein beliebiges Element P von ~ gilt

wobei r eine bestimmte mir von @ abhängige natürliche Zahl ist, die den folgenden Bedingungen ·genügen : r$1 mod. p,

3. In dieser Nr. soll die Struktur derjenigen M-Gruppe untersucht werden, deren ·Ordnung durch genau drei verschiedene Primzahlen teilbar ist. Im folgenden seien P1, P2, Pa immer verschiedene Primzahlen mit P1>P2>Pa,

Über die endlichen Gruppen und die Verbände ihrer Untergruppen.

179

Hilfssatz 11. @ sei eine M-Gruppe von der Ordnung pf1pq2pf3 , urid ~~. sei eine Pi-Sylowgruppe von @ (i=l, 2, 3). Wenn ~1\.;)~2 nilpotent ist, dann ist es auch entweder ~ 1 v ij3 3 oder ~2 v ~ 3• Beweis. Zunächst bemerken wir, dass ~"v ~; (i~j) eine Untergruppe von der Ordnung p~iP? ist. (Vgl. Hilfssatz 4). . Wenn ~ 1 v ~3 und l~2 0 ~s beide nicht nilpotent sind, so haben sie die im' Satz 2 angegebenen Struktur. Daher fällt der Normalisator .N(~3} von ~ 3 in@ mit ~s ·zusammen: N(~3) =~3• Pi bzw. P2 sei von 1 verschiedenes Element von ~1 bzw. ~2- oa·P1~g.P- 1 ~P2~a.P21 ist, so' gibt es nach Hilfssatz 10, ein in ~ 1 v ~ 2 enthaltenes Element. Q von einer Primzahlordnung; (~Pa), so dass QP1~aP{ 1Q-1 =P2~sP21 ·ist. Aus Nl~3)=~3 folgt dann QPi =P2, was aber wegen ~ 1 2 = ~1 x ~2 ein Widerspruch ist. Es muss also entweder ~ 1 v ~ 3 oder ~ 2 v ~ 3 nilpotent sein. Hilfssatz 12. @ sei eine M-Gruppe von der Ordnung pf1pq2pff3 und t~,: sei eine Pi-Sylowgruppe von @ .(i=l, 2; 3). Wenn ~ 1 v ~ 2 nicht nilpotent ist, dann sind ~1 v ~ 3 und ~2 v ?ß3 beide nilpotent, d. h. es gilt @= (~1 V ~2) X ~3• Beweis. Sei :zunächst ~ 1 u ~ 3 nicht nilpotent. Nach Satz 2 ist dann ~s zyklisch: ~ 3 = {P3}, und für jedes Element P1 von ~ 1 gilt PaPiP:;1 = 11., r1. :\:= 1 mod. Pi, dPa= 1 niod. P1 • Da nach Hilfssatz 5 der Normalisator N(~2) von ~ 2 jede p3-Sylow• gruppe von @ enthält, so liegt (Pi- 1PaPi)P:;1 =Pr- 1 in N(~2). ~i ist also in N(~2) enthalten. Da ~\ ein Normal teiler von: ~ 1 v ~ 2 ist (vgl. Hilfssatz 5), so folgt hieraus ~ 1 v~ 2 =~1 x ~ 2, was unserer Voraussetzung widerspricht. Daher muss ~ 1 v~3 nilpotent sein. Nun sei ~ 2 U~s wieder nicht nilpotent. Nach Satz 2 ist dann ~ 2 eine zyklische Gruppe von .der Ordnung 'P2: ~2= {A}. Wenn ~ 3 = {P8} gesetzt wird, so gilt für jedes Element P1 . von ~ 1

v~

(4)

P2PiP21 =Pf,

a. ~

(5)

PaP2Pi 1 =Pf,

ß:\:= 1 mod. P2, ßP =l mod. P2,

,(6)

P1Psl'11 =Ps.

1 mod. P1,

a.P2

=l mod. Pi,

3

Daraus folgt Pf 8 =Pf,

also

a8- 1 =1

mod. Pi, ß=l mod. P2,

was aber mit zweiter Bedingung von (5) in Widerspruch steht. Daher muss auch i2v~8. nilpotent sein, w. z. b. w.. Hilfssatz 13. @ sei eine M-Gruppe von der Ordnung pf 1pi2pff3, und

[3)

17

180

Kenkiti Iwasawa.

~i sei eine Pi·Sylowgruppe von eine direkte Zerlegung

@

(i=l, 2, 3).

Dann besteht immer

@=(~i1 V ~i2) X ~is,

wobei (i1, ~' is) eine geeignete Permutation von (1, 2, 3) ist. Beweis.. Wenn ~ 1 u ~ nilpotent' ist, so gilt nach Hilfssatz· 12 ent"' weder @== (~1 u ~a) x ~2, oder @= (~2 u ~s) x ~1. Wenn aber ~1 u ~2 nicht nilpotent ist, dann besteht nach Hilfssatz 12 auch eine direkte Zeregung @= (~i v ~2) x ~aUm die Struktur der allgemeinen endlichen M-Gruppe zu bestimmen, benutzen wir noch einen Hilfssatz : Hüfssatz 14. @ sei das direk1;e . Produkt von endlichen Gruppen @1, @2, ••• , @r, deren Ordnungen zueinander teilerfremd sind. Der zu @ gehörige Verband V(@) .ist dann die dire~te Summe ·der zu @ ge. hörigen. Verbande V(@i) (i=l, 2, ... , r). Wenn also ·@i sämtlich MGruppen sind, dann ist auch @ eine M-Gruppe. Beweis. Es ist leicht einzusehen, dass· eine beliebige Untergruppe U von @ das direkte Produkt von Ui=U f"'i@, (i=l, 2, .. ,, r) ist. Nun seinen & und 58 zwei beliebige Untergruppen von @. Aus

folgt dann

%f V 58 = (%f1 V 581) X (%f2 V 582) X · ·' %(

f"'i58=(%f1("\581)

X

(%f2("\582) X."'

X

(filr V 58r)

X

(&rf'"'\~r).

Und

Daraus folgt unmittelbar die Behauptung.. Das direkte ·Produkt zweier M-Gruppen ist· aber nicht ·immer eine M-Gruppe, wenn ihre Ordnungen nicht teilerfremd sind. Z. B. ist das folgende @ keine M-Gruppe, obwolil & und 58 beide" modular sind :

@=&x58, &={A,B}, 58={C},

AP2 =B11 =l., BAB-1 =A1+ 11 , Cpll=l,

(p>2).

Mit Hilfe der Hilfssatz 13, 14 kann man nun leicht die Struktur von endlichen M-Gruppen bestimmen. @ sei eine M-Gruppe von der Ordnung g=pf 1pf2 . . . p:r. Eine PiSylowgruppe von @ sei mit ~i bezeichnet (i=l, 2, ... , r). Es ist dann offenbar @=i1v~2V···V~r· Wenn ~1u~. (i=2, ... ,r) sämtlich nilpotent sind, so gilt @= ~ 1 x (~2 v ~ 3 u , .. u ~r)'. We~n aber z. B. lß1 v ~ 2 nicht nilpotent ist, dann ist nach Hilfssatz 13 ~1 v ~ 2 . mit jedem

[3]

18

Über die endlichen Gruppen und die Verbände ihrer Untergruppen.

181

(:j = 3, 4, ... , r) elementweise vertauschbar, also besteht eine direkte Zerlegung @= (~1 u s:ß2) X(s:ßa V s:ß4 V··· U s:ßr), Da (s:ß2 U !ßa V · · · u s:ßr) bzw. (?ß3u s:ß4u ·· · v ?ßr) wieder eine M-Gruppe ist, so kann man dieses Verfahren fortsetzen, bis endlich @ ins direkte Produkt einiger ?ßi und einiger ?ßiu !ß1c zerlegt wird. Dieses !ß" bzw. !ßjv s:ß1c ist offenbar modular, und da ausserdem s:ßiv ~1c nicht nilpotent ist, so besitzt dieses die im Satz 2 angegebene Struk_tur. Wenn umgekehrt eine Gruppe @ das direkte Produkt von denjenigen M-Gruppen ist, deren Ordnungen durch höchstens zwei verschiedene Primzahlen teilbar, und zueinander teilerfremd sind, dann ist nach Hilfssatz 14 @ selbst modular. Also haben wir Satz 3. Eine endliche Gruppe @ ist dann und nur dann modular, wenn sie ins direkte Produkt von einigen modularen.Sylowgruppen und einigen Gruppen mit der. im Satz 2 angegebenen Struktur zerlegbar i_st. Die direkten Faktoren sollen dabei zueinander teilerfremden Ordnungen besitzen. ~j

§2. Wir untersuchen nun die Struktur von endlichen D-Gruppen. 10> Da eine D-Gruppe offenbar zugleich eine M-Gruppe ist, so können wir alle Ergebnisse von § 1 brauchen. · Sei @ zunächst eine D-Gruppe von der Ordnung p2, wobei p wie immer eine Primzahl bedeutet. Wenn @ nicht zyklisch wäre, und @={A, B}, 2l={A}, ?B={B}, @:={AB} gesetzt wird, so wäre fil =fil v (?8 fl@:) =\= (fil u?B) fl (21 v@:)=@,

was unsrer Annahme. widerspricht. @ ist also zyklisch. Hieraus folgt nun ohne Weiteres, dass ·eine distributive p-Gruppe entweder zyklisch oder eine verallgemeinerte Quaternionengruppe ist, da ihre Untergruppen von den Ordnungen p 2 sämtlich zyklisch sind.m . Eine verallgemeinerte Quaternionengruppe ist. aber wie leicht ersichtlich nicht distributiv. Jede distributive p-Gruppe muss daher zyklisch. sein. @ sei nun eine DGruppe von der Ordnung paqß. Wenn@ nicht nilpotent wäre, so müsste es die im Satz 2 angegebene Struktur besitzen; @={P,Q}, PP=l,

r ~ 1 mod. p, Setzt man

m= {P},

78= {Q},

Qq'fl,=1

QPQ-1 =P",

rq= 1 mod. p • [ = {PQF- 1},

dann gilt

10) Vgl. O. Ore, Duke Math. Jour., 4, 247-269, 1938. 11) Vgl. H. Z., a. a. 0. S. 113.

19

[3) Kenkiti Iwasawa.

182

was wieder nicht geht. @ muss also nilpotent sein. Hieraus folgt nach Satz 3, dass eine endliche D-Gruppe das direktes Produkt ihrer zyklischen Sylowgruppen, also selbst zyklisch ist. Sei nun @ umgekehrt eine endliche zyklische Gruppe, und ~r. 58, Q:: ihre beliebige Untergruppen. Wenn die Ordnung von m, 58, Q:: bzw. mit a, b, c bezeichnet wird, so bedeutet die Bedingung

m· v

(>B "Q::) =(m u >B) n (& v Q::)

nicht anders als eine elementar-zahlentheoretische Relation {a, (b, c)} =({a, b}, {a, c}). @ ist also eine D-Gruppe. Damit ist der folgende Satz bewiesen. ·Satz 4. Eine endliche Gruppe @ ist dann und nur dann distributiv, wenn es eine zyklische Gruppe ist. Bevor wir die Struktur der endlichen B-Gruppen untersuchen, seien einige Bemerkungen über K-Gruppen vorausgeschickt. Die Struktur der allgemeinen endlichen K-Gruppen scheint sehr kompliziert zu sein. Z. B. ist eine K-Gruppe nicht immer auflösbar, und .die Untergruppe einer endlichen K-Gruppe ist nicht immer eine K-Gruppe. (Z. B. & 5 bzw. ®4). Eine K-Gruppe von einer Primzahlpotenzordnung ist aber leicht zu bestimmen. @ sei also eine komplementäre p-Gruppe und 1 r-. 58 bzw. a, b, d ist, so ist die Anzahl der Elemente von fil58 gleich ab/d. Da 2C mit 58 dann und nur dann vertauschbar ist, wenn s.>(58 =~l u 5,5 gilt, so folgt hieraus Hilfssatz 15. Die Untergruppen &, 58 einer endlichen Gruppe ist dann und nur dann vertauschbar, wenn ab=dm gilt, wobei a, b, d, m bzw. die Ordnung von fü, 'i8, 2C n 58, 2( v 'i8 bedeutet. Nach 0. Ore15 > nennen wir eine Untergruppe fil von @ ,, quasinormal ", wenn fil mit jeder Untergruppe 58 von @ vertauschbar ist. Wir nennen dann eine Gruppe @ ,, quasi-Hamiltonsch ", wenn jede Untergruppe von @ quasi-normal ist. Es ist klar, dass ein Normalteiler quasi-normal ist, und folglich eine abelsche Gruppe bzw. eine Hamiltonsche Gruppe immer quasi-Hamiltonsch ist. Wenn nun 2C mit 'i8 vertauschbar ist, so gilt, wie bekannt ist und übrigens leicht zu sehen ist, für jedes a:::::, fil die Relation

Eine quasi-Hamiltonsche Gruppe ist daher stets modular. Nach Hilfssatz 6 und Hilfssatz 15 ist umgekehrt eine modulare p-Gruppe auch quasi-Hamiltonsch. Also gilt Hilfssatz 16. Eine p-Gruppe @ ist dann und nur dann quasiHamiltonsch, wenn es modular ist. Für allgemeine endliche Gruppe @ gilt nun Satz 7. Eine endliche Gruppe @ ist dann und nur dann quasiHamiltonsch, wenn sie das direkte Produkt ihrer modularen Sylowgruppen ist. Beweis. Es sei @ eine quasi-Hamiltonsche Gruppe und ~ 1, ~z. seien ihre p-Sylowgruppen. Die Ordnung von ~ 1 u ~ 2 ist dann. nach Hilfssatz 15 wieder eine p-Potenz, also ~ 1 =~2 =~ 1 v~2. @ist also nilpotent und aus Hilfssatz 16 folgt die Behauptung. Die Umkehrung ist auch nach Hilfssatz 14, 16 klar. Nach diesem Satz ist auch die Untersuchung von endlichen quasiHamiltonschen Gruppen auf die von modularen p-Gruppen zurückgeführt werden. Nach Satz 3, 7 folgt nun ohne Weiteres 15) 0. Ore, Duke Math. J. 3, 149-174, 1936.

Über die endlichen Gruppen und die Verbände ihrer Untergruppen.

185

Satz 8. Eine endliche M-Gruppe @ ist dann und nur dann quasiHamiltonsch, wenn sie nilpotent ist. Es gibt aber zufolge Satz 2 nicht nilpotente M-Gruppen ; eine M-Gruppe ist also nicht immer quasi-Hamiltonsch. Zwei Untergruppen einer endlichen M-Gruppe sind aber nach Hilfssatz 4 stets miteinander vertauschbar, wenn ihre Ordnungen zueinander teilerfremd sind. Allgemeiner gilt nun Satz 9. @ sei eine endliche M-Gruppe und m, 58 seien beliebige Untergruppen von@. Es gilt dann stets eine konjugierte Untergruppe von 2r, die mit ~ vertauschbar ist. Beweis. Nach Satz 3 und Hilfssatz 16 genügt es den Satz nur für die Gruppe mit der im Satz 2 angegebenen Struktur zu beweisen. Man wähle alsdann eine konjugierte Untergruppe x21x- 1 von ?lT, so dass (in derselben Bezeichnungsweise wie im Satz 2) eine q-Sylowgruppe von xmx- 1 und eine q-Sylowgruppe von 58 in einer und derselben qSylowgruppe von @ enthalten werden. Dann ist offenbar x21x- 1 mit 58 vertauschbar. Aus diesem Beweis ist leicht einzusehen, dass die Konjugierte von ~ sogar in 21" u 58 gewählt werden kann. Diese Bedingung· ist aber nicht hinreichend dafür, dass eine endliche Gruppe modular ist. Z. B. ist die folgende Gruppe @ keine M-Gruppe, obgleich es die genannte Eigenschaft besitzt, (vgl. den Beweis von Hilfssatz 8) : @= {A,

B}

X

{C},

AP=l, Bq=Cq=l, BAB- 1 =Ar,

r~ 1 mod. p, Für eine p-Gruppe ist jedoch auch diese .Eigenschaft hinreichend. Vgl. Satz 12 in § 4. Im Anschluss an Obigem soll ein Problem von 0. Ore16' betrachtet werden. Wenn zwei Untergruppen 2!, 58 einer Gruppe G mit einander vertauschbar sind, so gilt für beliebige Untergruppen mit (r :::> & bzw. '.!l :::> 58 von @ die Relationen ((r :::im), (8) (21" 0 58) t') (r = ~r u (58 t') (r) ( '.!) :::> 58) • (9) ()Bu 21") n :!l=58 v (~I t') :!l) 0. Ore hat die Frage aufgeworfen, ob umgekehrt die Relationen (8), (9) dafür hinr~ichend sind, dass & mit 58 vertauschbar. ist. Dass (8) und (9) dafür nicbt immer hinreichend sind, ergibt sich sofort daraus, dass es eine nicht quasi-Hamiltonsche M-Gruppe gibt. Nun gilt 16) Vgl. 0. Ore, a. a. 0. 11).

[3]

23 Kenkiti Iwasawa.

186

Satz 10.17> Dafür, dass zwei beliebige Untergruppen fil, 58 einer endlichen Gruppe G immer miteinander vertauschbar sind, 59bald sie die Bedingungen (8) µnd (9) erfüllen, ist notwendig und hinreichend, dass @ nilpotent ist. Beweis, Wir nehmen zunächst an, dass zwei beliebige·Untergruppen fil, ~ von.@ immer vertauschbar sind, wenn sie (8) und (9) erfüllen. g sei die Ordnung von @, Ull,d p ein Primteiler von g. Vorausgesetzt, dass eine p-Sylowgruppe ~ kein Normalteiler von @ ist, bezeichnen wir eine den Normalisator N(~) von ~ enthaltende maximale Untergruppe von @ mit im. (Evtl. 9R=N(~)). Die Ordnung von ~. N(~) sei bzw. pa, n und der Index [@ : Wl-], .[illi : N(~)] sei bzw. l, m. Da nach Sy1owsatz der NormaHsator N(ffn) von im mit lOl selbst zusammenfällt, so enthält @ genau l konjugierte Untergruppen von im, und von lm pSylowgruppen von @ sind je m in diesen Konjugierten enthalten. Für im und eine Konjugierte 'im1 von IDl gilt, wie leicht ersichtlich, die Relationen (8), (9). Nach Voraussetzung ist dann 9R mit im1 vertauschbar, also gilt zufolge Hilfssatz 15 gd=(mn)2, oder ld=mn, wobei d die Ordnung von 9R f""1 im1 bedeutet. Aus pa l n und pa-\-" l folgt nun pa Id. ·'im und 9R1 müssen also eine p-Sylowgruppe von @ gemeinsam enthalte1,1, was aber oben erwähnter Tatsache widerspricht. -~ ist also ein Normalteiler von @. Da p ein beliebiger Primteiler von fJ war, so folgt hieraus dass @ nilpotent ist. Sei nun @ umgekehrt eine endliche nilpotente Gruppe und fil, ?13 zwei Untergruppen von @, die (8) und (9) erfüllen. Wenn ~=~1

x ~2 x ··· x ~r und entsprechend fil =&1 x.ih x ··· x fil.,,

58 =?131 x ?132 x · · ··x ?13„ gesetzt. wird, dann ist fil X?l3=(fil1X?l31) x (fil1 X581) x ··· x.(~T,.>; für p > 2 sind sie nämlich durch die Relationen gegeben: (10) Apl'=BP=l, BAB- 1 =A1+P, (11) AP=BP=(A, B)11 =1, A(A, B)=(A, B)A, B(A, B)=(A,B)B, und für p = 2 durch (12) die Quaternionengruppe: A 4 =1, A 2 =B'-, BAB- 1=A-1, (13) die Diedergruppe: A 4 =1, B2 =1, BAB-1 =A'-1• Es. ist nun leicht nachweisbar, dass unter diesen Gruppen die Typen (10) und (12) modular, die Typen (11) und (13) aber nicht.. modular sind. Es gilt also Satz 1-1. Eine nichtabelsche Gruppe @ mit der Ordnung p3 ist dann und nur dann modular, wenn es den Typus (10) bzw. (12) hat. Folgender Hilfssatz ist nun oft nützlich, wenn man eine p-Gruppe als modular nachweisen will. Hilfssatz 17. @ sei eine p-Gruppe, deren echte Untergruppen sämtlich modular sind. Wenn @ selbst doch keine M-Gruppe .ist, so enthält @ einen Normalteiler in, so dass @/fil eine Gruppe vom Typus (11) oder (13) ist. Beweis. Da V(@) nach Voraussetzung kein modularer Verband ist, so gibt es nach einem bekannten Satz der Verbandstheorie19> solche Untergruppen 2(, ~ von @, dass

r

Uu~=@, .[@:mJ>p2, [@:~J>r, [&:&"~]=[~:~1'1'8]=p gilt. Ferner sei & 1 bzw. ~ 1 eine & bzw. ·~ enthaltende Unterguppe von@ mit dem Index [@: ~J'=p bzw. [@: ~J=p. in=&"~ ist nun offenbar ein Normalteiler von @. Wir setzen also @=@/in, fil=&/m, i3=~/m, fil1=&1/m, ~1=~1/m. 18) Vgl. z.B. H. Z., a. a. 0. S. 114. 19) Vgl. G. Birkhoff, a. a. 0. 8).

[3]

25

188

Kenkiti Iwasawa.

Wenn ferner

fil 2 ={Ä,BÄB-1,B2ÄB-2 ,

fil={Ä}, ~={B},

••• }

gesetzt wird, so ist &2 ein Normalteiler von ·~ mit fil C: ~2 C:: §11, fil2u~=~, i2f"'l~==l.

Daraus folgt [@ : mJ =p, also 212 = &1 = {A, BAB-1, B2AB- , ... }. Genau so erhalt man ~ 1 ={.B, ÄBÄ- 1, Ä2BÄ- 2, ••• }. Da nach Voraussetzung & 1, ~1 modular sind, so sind· &i, ~1 zufolge Hilfssatz 7 beide elementare abelsche Gruppen.20' Wenn die Ordnung von fil 1 pf 'ist, und wenn -

-

-

-

X. =liJ{B-1 ,

-

---

---2

für

11

gesetzt wird, so können wir [(1 als !-dimensionalen Vektorraum über GF{p), und " als lineare Transformation in diesem Rauin auffassen, lindausfil1 ={.A,Ä11,Ä 2. Dass umgekehrt ·jede solche 2-Gruppe modular ist, kann man genau so wie im Fall p > 2 beweisen. Mit Hilfe von Satz 13, 14 lässt sich nun folgenden Satz nach einigen Rechnungen ohne Schwierigkeit beweisen. Satz 15. Die nichtabelsche M-Gruppen von der Ordnung p4 haben folgende Typ~n : (1) für p>2, (15) A 118 =BP=l, BAB- 1 =Al+P\ {16) Apl=BP=C"=l, BAB-1 =Al+P, AC=CA, BC=CB, (17) A 111 =Bp3=l, BAB-1 =A1+ 11 , (2) für p=2, · 8 (18) A =B2=1~ BAB- 1=A6, (19) A4 =C2=1, A 2 =W, BAB-1 =A- 1, AC=CA, BC=CB. Es sei nun bemerkt, dass für zwei beliebige Elemente C, D in einer Gruppe @ vom Typus (16) oder (17), die Gleichung (CD) 11 =C11D 11

gilt, was später benutzt werden wird. HüfSBatz 18. Es [email protected] modulare p-Gruppe·und p1A die maximale Ordnung. eines Elementes aus @. Diejenige Elemente · aus @, deren p4 -te Potenzen gleich. 1 sind, bilden eine charakteristische Untergruppe !Ja von @,23> · ·Es gilt dann . 1 = JloC. .!21 C. ··· C: .JJJA-1 c. .!2µ= G, 22) Vgl, z.B.. W. Bumside, a.a..O. 5), S. 126. 23) Wir folgen die Bezeichnungen von l?. Hall, Proc. London Math. Soc. (2) 36, 29-96, 1938.

Über die endlichen Gruppen und die Verbände ihrer Untergruppen.

191

und f2,t/12i-i (i=l, 2, ... , µ) sind sämtlich elementare abelsche Gruppen. Beweis. Alle Elemente X aus @ mit x.'P = 1 bilden nach Hilfssatz 7 eine charakteristische Untergruppe 121 von@. In @=@/121 bilden die Elemente X mit XP= 1 wieder eine. charakteristische Untergruppe Ji1• Die in @ li1 entsprechende Gl°\1ppe !22 besteht offenbar genau aus allen Elementen Y von @ mit YP2 =l, und ii1=122/121 ist eine elementare abelsche Gruppe. In derselben Weise kann man der Reihe nach. die Gruppen .Qs, 124, . •• bilden und. beweisen, dass JJ.d 12.H alle elementar abelsch sind. Im folgenden sollen µ=µ(@) und 12a. immer den in diesem Hilfssatz angegebenen Sinn haben. @ sei nun eine modulare 2-Gruppe mit _µ(@)=2, und. enthalte eine Quaternionengruppe .0 : D={A,B},

A4 =1,

A 2=B2, BAB· 1 =A-1 •

Wir nehmen zunächst an, dass [@ : !J1] > 28 ist, und bezeichen mit C ein nicht in 0,!21 entpaltenes Element von@. Wenn C 2 =A2 =-B2 ist, dann ist {A, B, C} eine Gruppe von der Ordnung 24, andererseits ist sie aber nicht vom Typus (19), was unmöglich ist. Es ist also {C} ~ {A, B} =1, .und nach Satz 15 sind {A, C}, {B, C} beide abeisch; Da @ nach Hilfssatz 16 quasi-Hamiltonsch ist, so gibt es ganze Zahlen x, y, so :dass B(AC)=(AC)"'B11 gilt. Nun ist B(AC)=4- 1BC=(AC)"'B1'=A"'C"'B11 =A:uB11C"',

also folg}tch x=l, y= -1 mod. 4, · (AC)-1B(AC)=B- 1 •

Da B2 nicht mit (AQ) 2 =A2C2 zusammenfällt, und im Zentrum von {(AC), B} enthalten ist, so ist {AC), B} / {B2} e.ine Diedergruppe, also nicht modular. .ln unserem Fall ist das.nicht möglich; es muSß @=0121, namlich [@: !J1]=2 sein. Wenn. nun A 2, C1, C2, ... , C„ eine Basis von JJ1 ist, so sind {Ci, A}, {Ci, B} nach Satz 11 beide abelsch. . @ ist mithin . das direkte Produkt von einer Qua.ternionengruppe und einer abelschen Gruppe vom Typus (2, 2, ... , 2). Dass umgekehrt eine solche Gruppe wirklich e.ine M-Gruppe ist, kann man l~icht beweisen. Nun sei ·@ eine modulare 2-Gr.uppe mit µ(@) = 2, die keine Quaternionen,-

29

[3] 192

Kenkiti I wasawa.

gruppe enthält, und C, D beliebige Elemente aus: @. Die Ordnung von { C, D} ist dann offenbar hochstens 24• Da es aber keine Quaternionengruppe enthält, so folgt aus Satzen 11, 15 dass {C, D} abelsch ist. @ selbst muss daher abelsch sein. Es gilt also Hilfssatz 19. Eine modulare 2-Gruppe mit µ(@)=2 ist entweder abelsch oder direktes Produkt von einer . Quaternionengruppe und einer abelschen Gruppe vom Typus (2, 2, ... , 2). Die letztere Gruppe wollen wir die spezielle 2-Gruppe nennen. 24> Nunmehr sei @ eine allgemeine modulare 2-Gruppe. CYJ=!Jl!2i-2 (i=2, 3, ... , µ) ist dabei natürlich eine modulare 2-Gruppe mit µ(G,)=2. Wir nehmen nun an, dass für einen Index i !Ji/,Qi-2 nichtabelsch ist, dass also Qi/ !2.-2 eine spezielle 2~Gruppe ist. Sei zunächst i < µ. Falls die in 2) mit µ(@)=2, und C,D beliebige Elemente von @. Die Gruppe {C, D} ist dann entweder 24) Eine spezielle 2·Gruppe ist also nicht anders als eine Hamiltonsche 2·Gruppe.

Über die endlichen Gruppen, und die Ve.rbände ihrer Untergruppen.

193

abelsch oder vom Typus (10), (16), bzw. (17). Es gilt also, wie schon bemerkt, (20) (CD)P= CPDP. (20) ist_ auch für eine modulare 2-Gruppe @ mit µ(@) = 2 gültig, wenn es keine spezielle 2-Gruppe ist. Es gilt nun Hüfssatz 20. Es sei@ eine modulare p-Gruppe (p>2) mitµ(@)=µ, die keine spezielle 2-Gruppe ist. Für beliebige Elemente C, 1J aus @ gilt dann die gleichung (CD)P1,1-1 =cz1A-1npµ ...1. Beweis. @=!Jµ/!2µ-2 ist eine modulare p-Gruppe .mit µ(@)=2. Es gilt also nach (20) d. h.

(CD)P=CPDPZ,

Z e !21,1-2.

Alle Elemente in dieser Gleichung sind in .!J1,1-1 enthalten. Wenn man also in !Jµ-1/!21,1-a rechnet, so folgt daraus wied~r nach (20)

(CD)p2=CP2DP 2ZP=CP 2DP1 (f2µ_ 8),. d. h. So fortfahrend. erhält man schliesslich

(CD)Pµ-1 =C'llµ-lDPµ-1 ' w. z. b. w•• Sei @ .wieder eine modulare G.ruppe mit der Ordnung pn, die keine spezielle 2-Gruppe · ist. Nun setzen wir [!211 : !.?11-1]=p"'11 (a=; 1, 2, ... , µ), Wir wählen Elemente· A1, A2, ' ...' ,' A"'µ die in !.?µ{!21,1-1 . eine ' Basis bild.en. Indem man Hilfssatz 20 auf !.?1,1/JJ1,1-2 anwendet, sieht man leicht ein, dass Af, Af, ... , A:µ zu einer Basis Af, Al, ... , A!µ, A"'µ+1, ... , A"'µ-t von JJ1,1:-1/JJ1,1-2 ergänzt werden kann. Dies~ Eleznentensystem lässt sich dann wieder zu einer .Basis von .Qµ_JJJ,,-a ergänzt werden~ und so fortfahrend kann man schliesslich ein Erzeugendensystem Ai, ... ~ A"'p. , ... , A(dµ- 1, ••• , A...1 von @ erhalten. Ist die Ordnung von Ai gleich p\ so ist jedes Element aus @ in der Form (21)

X=Af1Af2 •••

~ ~1 ,

O~x,a erfüllen, wird eben durch [JJa: .Q4 _1]=p"'a angegeben. Insbesondere gil~ a1 = w1, Wir bestimmen nun die Struktur der modularen p-Gruppe mit w1 2) sämtlich kleiner als pJJ ist. Falls µ-1 < s ist, so ist @' = 1 oder @' =,8. Nach Hilfssatz 22 sind wir fertig. Wenn aber µ -1 > s ist, so ersetzen wir T durch T * = T 1- hPµ-1-s • Es gilt dann wieder T* A,T*- 1 = A1+pa (i= 1, 2, ... , r). b) Sei nun ,8 E {T}. Wir wählen eine Basis A 1, ... , A,., so dass {T} nm 1 in {A1} enthalten ist. Die Ordnung von T sei p... Wenn die Ordnung eines Ai nicht kleiner als p"" ist, so kann man T durch T* = T At ersetzen und den Beweis auf dem Fall a) zurückführen. Wir nehmen daher an, dass n=µ ·ist, und die Ordnung von Ai (i=l, 2, ... , r) sämtlich höchstens p/J,- 1 ist. Es ist nun leicht ·ersichtlich, dass man für i>3 TA.T- 1 =A}+p• und für i=2 TA 2T- 1 =Al+ps oder TA 2T- 1 = Ai+PsTp1.1-i ohne Beschränkung der Allgemeinheit annehmen kann. b1) Es sei TA 1T- 1 ==Af+Ps+hi;,""- 1, TA,T- 1 =AJ+P3 (i>2). Wenn man ein geeignetes o. wählt und A1 durch Al= A 1Ta ersetzt, dann wird A{, A2, ... , Ar offenbar wieder abelsch und es gilt TA;T:- 1=A: 1+p8, TAiT- 1 =A1+ps (i> 2). ~) Sei nun TA1T- 1 =Af+P3 +kpm-l, TA2T- 1 =A~+P3 TPµ-l' TAir- 1 = AtPs (i 2; 3). Wenn s < µ -1 ist, so ersetzte man A2 durch At= A 2TPµ-I-s, und reduziere den Beweis auf dem Fall bi), Wenn aber s> µ-1 ist, dann ist entweder @'=l oder @'=,8. Nach Hilfssatz 22 sind wir auch dann fertig. (II) mi=,8. Nach Hilfssatz 22 gilt

m1={A1, A2}

X

{As, A4, ... ,Ar}, Afi=l, AiAi=A3A,: (i,j> 3).

Ausserdem kann man noch annehmen, dass TA 1T- 1 =Ai+P8, TA2T- 1= Al+Ps ist, indem man evtl. T durch T* = TAtAf ersetzt. Z ist nun

PJ

36 Über die endlichen Gruppen und die Verbände ihrer Uptergruppen.

nach Definition die p"-1-te Potenz eines Elementes von @. Z=Zt· 1, Zo=Af 1A~2 ••• A;rrr. Nach Hilfssatz 20 gilt dann A lpm-1 -_Z-A•1Pµ-lAe2Pµ-lTfPµ-l l 2 m-1

199

Sei .also



µ-1

Wenn also Af = 1, so muss TP =\= l sein. a) Sei zunächst Afm- 1 =\=l. Es gilt dann offenbar µ=m. Wenn B µ ist, so ist @' =.8 und sind wir fertig. Sei s =µ -1. Wenn At'-1 = 1. ist, dann ist wieder @' = ,8. Es sei also A Pµ-t =\= 1. · Nach einiger Rechnung folgt dann TAiT-'- 1 =Ai (i>3) und man kann den Beweis auf dem Fall (I) reduzieren, indem man A 2 durch · A; = A 2 1 ersetzt. Sei endlich s < µ -1. Der Beweis wird wieder auf dem Fall (I) zurückgeführt, indem man A 2 durch A; =A2T-i>"'-l-s ersetzt.

z

r-

b) Es sei nun At- 1 =1, (m 0, i =

1, 2, ... , n.

If one denotes the commutator subgroup of G by G', then G /G' is an abelian group of order e = e 1 e2 ... en, On the other hand, it can easily be proved by a method of Schreier that G' is a free group. It follows that the intersection of all subgroups with finite index in G is equal to 1. But, in general, this property of G is equivalent to saying that the intersection of all normal subgroups with finite index in G is equal to 1 (Lemma 4). Thus one has

Theorem 7. Let G be a group generated by S1, S2, ... , Sn with the relation (4). Then the intersection of all normal subgroups of G with finite index is equal to 1. In particular, G is maximally almost periodic, · (The special case of Sl = 1, Sf = 1, for which the group G is the classical modular group, was given by J. von Neumann as an example of a m.a.p. group.) Next we consider a group with the relations

(5)

Sf1

= 8 22 = ... = 8!", ei > O, i = 1, 2,

... , n

with n 2:: 2. By Theorem 6, if N1 and N2 are normal subgroups of G with N1 nN2 = 1 and if both G/N1 and G/N2 are m.a.p., then so is also G (Lemma 5). Now letG be a group with the relation (5) and set Z = Sf 1 8 22 = ... = 8!". Let G' be the commutator subgroup of G and put G'{Z} = G1, G' n {Z} G2, Then, since n ~ 2, it can easily be shown that [G 1 : G'l =[{ Z} : G2} is infinite, and so G2 = 1. Since by Theorem 7 both G/{Z} and G/G' are m.a.p., so is also G. This proves

=

=

Theorem 8. A group G generated by 81, 82, ... , Sn with the relation (5) is maximally almost· periodic. 5. Modifying V. L. Nisnewitsch's method (Rec. Math. 51 (1940)), one can show that a free product of any number of m.a.p. groups is also m.a.p. (Theorem 9). It follows that in Theorems 7, 8 the conditions ei > 0 may be replaced by ei ~ 0 and also the number of generators n can be infinite. Combining the above results, one can show that there are many other groups which are m.a.p. However, according to Fuchs-Rabinowitsch (Comptes Rendus

[ 4 ]

39

( D o k l a d y ) ,2 7 ,29( 1 9 4 0 ) ) ,t h eg r o u pw i t ht h er e l a t i o n s

S心 =8 迅, s ?=1 , s函 =S 心S a , S ふ = S?S4 i sn o tm . a .p .I ts e e m sani n t e r邸 t i n gp r o b l e mt oc o n s i d e ru n d e rwhatc o n d i t i o n s t h eg r o u p so fs i m i l a rt y p e sa . r em . a . p .

概週期函敷に就いて 東京帝*岩澤 d i s c r e t e な瀞の上に於ける概週期函徽I こついて,

健吉

特に位相幾何學によく現ほれる二三の群が,所講

[ 4 ]

40

57

寄書概週期函敷に就いて

"maximally a l m o s tp e r i o d i c " であることを述べ

あ ,l 'ダ裕.....,ふk•

i1(G) in our notation) for a locally compact separable group G and studied its various interesting properties, particularly when the group G is compact or abelian .. Replacing R(G) by L( G) = £(l,P)(G) to be defined below, we study relationships between representations of G and those of L(G) (and R(G)), and generalize Segal's theorem ([SJ, Th.3). Though the details of Segal's proof are not available to us, our methods are probably similar to his. We use the notion of a normed ring in the sense of Gelfand (G] which has been used for studies of abelian groups by Gelfand-Raikov. In the first chapter of (5) we study a normed ring R satisfying the following assumption: (A) For any left [resp. right] maximal ideal 11 (resp. Ir) of R, the quotient R/11 [resp. R/Ir] is always finite dimensional. We obtain some results (Theorem 1-8) on the ideal structure of such a normed ring R, which are useful in the second chapter. For instance, we show {Theorem 8) that, if R satisfies the condition {A), then the "radical" of R (i.e., the intersection of all maximal two-sided ideals of R) is the largest left (or right, hence two-sided} ideal contained in the set of all "generalized nilpotent elements" in R (i.e., those = 0). elements x in R with lim Now, let G be a locally compact separable group andµ a right invariant Haar measure of G. We study £( 1 ,P)(G) = L 1 (G) n l,P(G} for p ~1. With the convolution product it has a ring structure. We denote by R(l,P)(G) the ring obtained from £(l,P)(G) by adjoining a multiplicative unit. For representations by finite dimensional matrices, we obtain the following main theorem (Theorem 9).

v'iixn1!

Theorem. We have a bijective correspondence between continuous representations of £(l,P)(G) and bounded continuous representations of G. That is, for a given continuous representation x(g) 1-+ T(x) of £Cl,P)(G), we have a unique bounded continuous representation a 1-+ DT(a) of G such that we have T(x)

=

fa x(g)Dr(g)dg

for any x(g) E £(l,P)(G). Moreover, for a given bounded continuous representation a 1-+ D(a) of G, the above integral formula gives a continuous representation x(g) 1-+ Tv(x) of £(l,P)(G). These constructions are the inverse to each other and preserve the equivalence relation of representations. As corollaries (of the proof), we obtain the following two results (Theorems 10, 11).

44

45

[ 5 ] C o r o l l a r y .Anyboundedm e a s u r a b l er e p r e s e n t a t i o no fG i sc o n t i n u o u s .

C o r o l l a r y .Anyc o n t i n u o u sr e p r e s e n t a t i o no f £ ( l , P ) ( G )i sc o m p l e t e l yr e d u c i b l e , andc o n v e r s e l y ,anyc o m p l e t e l yr e d u c i b l er e p r e s e n t a t i o no fL( l , p )( G )i sc o n t i n u o u s . Weh a v es i m i l a rr e s u l t sf o rR ( l , p )( G )(Theorems1 5 ,1 6 ) ,w h i c hi m p l yag e n e r a l i z a t i o no fS e g a l ' stheorem(Theorem1 8 ) .Asana p p l i c a t i o no ft h ea b o v em a i : r ; i . 9 ) . r e s u l t ,wea l s oh a v et h ef o l l o w i n gtheorem(Theorem1 Theorem.I fG i sn o tc o m p a c t ,t h e nwehavenon o n z e r or e p r e s e n t a t i o no fG b e l o n g i n gt oび ( G ) ,p; ? ; :1 .I fG i sc o m p a c t ,t h e nanyr e p r e s e n t a t i o nb e l o n g i n g t oLP( G ) ,p>1 i o n t m u o u s . sboun de dandc

References [ G JI .M.G e l f a n d ,N o r m i e r t eR i n g e ,R e c .M a t h .51( 1 9 4 1 ) ,3 2 4 . [ S JI .E .S e g a l ,Theg r o u pr i n go fal o c a l l yc o m p a c tg r o u p ,I ,P r o c .N a t .A c a d .S c i . , U . S . A . ,27( 1 9 4 1 ) ,3 4 8 3 5 2 .

1 0 8 8 . ノルム環卜 Segalノ定理ニツイテ

I

I . E .S e g a lハサキニ任意の l o c a l l yc o m p a c t ,s e p a r a b l eナ群 G 二封シ g r o u pr i n g R(G)ヲ定義シ,特二 G ガ a b e l群又ハ compact群ナル場合二種々ノ興味アル結果 ) 、 ノ ノ c ompact群二関スル主ナ結果ハ深宮氏ニョッテ簡輩ナ証明ガ輿 ヲ得マシタ.1 ヘラレマシタ.2 ) ココデハ G ガ一般二 l o c a l l ycompactナル場合ヲ考へ,特二 G ノ表現卜 R(G)ノ 表現トノ関係ヲ述ベテ見ヨウト思ヒマス.ソレハ S e g a lノ論文二於ケル定理 3ノ拡張 二嘗ルワケデスガ我々ハ R (G)ヨリモムシロ L(G)ヲ目標ニシテ進ミマシタ(後述). 然シ S e g a lノ証明ハ知ル由モアリマセンガ恐ラク以下述ベルモノト大同小異ナモノデ アリマセウカラ,以下ノ結果モ亦既知ノモノト思ハレマス.

I .ノルム環二関スル注意 §1.我々ハ G e l f a n dノノルム環 3 ) ヲ用ヒテ G e l f a n d ,R a i k o v 4 lガ a b e l群ノ場合ニ 用と夕方法二従フコトニシマスノデ,始メニ一般ノ非可換ナノルム環ニツイテ若干ノ 簡鞘ナ注意ヲ述ベルコトニシマス.勿論非可換ノルム環ノ一般論卜云フヤウナモノデ ハアリマセン. 先ヅ一般ノノルム環 R ヲ G e l f a n d ,N .R . §1ニヨリ定義シマス.但シコノ際乗 法ノ可換性ハ俵定シマセン.可換性ハナクトモ N .R .ニ於ケル代数的ナ定理ノ多クハ 成立スルコトガ容易二確カメラレマス.唯逆元 x-1 卜云フトキ左逆元卜右逆元トヲ区 別セネバナラヌコト,及ビ I d e a lニツイテモ左I d e a l ,右 ーI d e a l , 両側 I d e a lノ別ヲ考

4 6

( 5 )

へネバナラヌコト等ヲ注意スレバヨイワケデ,ソノヤウナ点ヲ考慮二入レサヘスレバ

N .R .ノ §1-§5ノ定理ハ凡テソノママ成立シマス. 特二

I tヲ任意ノ閉ヂタ左ーI d e a l トスレバ R / I tハ一般二環デハアリマセンガ兎モ

角B a n a c h空間デ R ノ各元 xハソノ(左カラノ)

O p e r a t o r 卜考ヘラレ XER , Y ER / l t二封シ l l x Y I Is ;l l x l l l l Y I I トナリマス. ( N .R . ,H i l f s s a t z4 )然シ可換ナ 場合卜本質的二異ナル点ハ輩純ノルム環ガ必ズシモ複素数体(或ハソノ上ノ有限次ノ

A l g e b r a ) トナラヌコトデ,コノクメ R ヲ m a x i m a lI d e a lデ割ッタ R/1ノ様子ガワ カラヌコトデス.

§2. ョッテ我々ハ先ヅ一番簡輩ナ有限次元ノノルム環ヲ調ベテ見ルコトニシマス. 複素数体

K 上ノ n次ノ全行列環ヲ Kn トカクコトニシ

L < l i , 但,j , a i , jEK, e i , iハ行列ノ鼠位 i , j

XEK n , X= 二射シテ

『 ( 苔 囮 ,,1')l

( 1 )

l l x

トオケバコノ n ormニヨリ Knガノルム環トナルコトハ明カデス.(但シ l l e l l *=1ニ ormニヨリノルム環トナ ハナッテヰナイ.)ヨッテ Knノ部分環ハ矢張リスベテ上ノ n リマスガ実ハ有限次元ノノルム環ハコレ以外ニハ存在シナイコトガ証明サレマス.ソ ノ為二先ヅ 補助定理

1 . s ) M ヲーツノ有限 K 加群トシソノ B函 sヲ u 1 , u 2 ,. . ., u n トスル.

B ハ Banach 空間デ K—加群トシテ M 二含マレテヰルモノトスレバ, B 二於ケル

norml l x l lハ M 二於テ定義サレル norm : n U n ! I "=( l o : 1 1 2十・・・+lo:nl2)½, l l x l l *=l l o : 1 U 1十・・・十 O ト等値デアル.即チ適嘗二 C1,C2>0ヲトレバ

O : iE K

B 二属スル任意ノ”二対シテ

C1llxll*~llxll~C21lxll* トナル. 証明 .B=Mノトキハ

l l x l l$ l a 1 II I 叫+・・ ・+l a n ll l u n l l

$ ( 名 『 ( 孟 1 ° , 1

C叫1•11'

l l u , 1 1 ' )I-

ナル故 l l x l l *カラ l l x l lヘノ寓像ハ連続トナリコレハ明白デス. 一般ニハ B ノ K B a s i sv i ,v 2 ,. . ., Vmヲトリコレヲ補充シテ Mノ B邸 i sV t ,虹 … , 髯 W t , , , ,,Wn-mヲツクリマス.

xEM,x=叩 1+・・・+叫n=馳+・・・+f 3 m V m+f 3 m + 1叫+・・.+( 3 匹 n -m

41

[ 5 ] ナルトキ

l l x l l * *= ( 1 ! 3 1 1 2十・・ ・ +lf3ml2+lf3m+l1 2十・・ ・ +lf3nl2)½

トオケバ B = Mナルトキノ結果ニヨリ I I 叫I * 卜 佃I I * * トハ等値.特ニェガ B ニ 3 m + l= ・・ ・=ふ =0 デ l l x l l * *ハ B 二於ケルーツノ normヲ与 含マレテヰレバ f I ト等値. ヘマスガ,再ビ B = Mナルトキノ結果ヲ用ヒレバコレハ B 二 於 ケ ル 肱I ヨッテ結局 norml l x l lハ B 内二 i n d u c eサレタ norml l x l l * 卜等値ニナリ定理ハ証明 サレマシタ. サテ任意ノ有限次元ノノルム環ハ環トシテハ勿論行列環ノ部分環トナリマスカラ上 ノ補助定理ニヨリ直チニ次ノ定理ガ得ラレマス.

.有限次元ノノルム環ハ ( 1 )ナル norm二関スル全行列環 Knノアル部分環 定理 1 ト一致スル. コノ定理ニョリ有限次元ノノルム環ノ様子ハワカッタワケデスガ,尚後デ用ビル定 理ヲココデ証明シテオキマス. 定理 2 .R=Kn二於テ ハ零巾元ナルコトデアル.

V戸可→ 0(n→oo)ナル為メニ必要且ツ十分ナル條件

証明.十分ノ方ハ明白デス.ョッテェガ零巾デナイトスレバ適当二 yヲトリ標準型

\ー︸

p



*n

P i z-yxy-'-(・..

ヲツクッタトキ少クトモーツノ P iハ 0デハアリマセン.コレヲ P k: /0 トスレバ H z 叶I *2 :IPkln• ョ ッ テ 血 乍可*2 :I P k l>0 . 即チ l i m切戸干.1 ' o . 任意ノ normハ l ! x l l * 卜等値ナル故 l i mv ' 匝可 1'0. ョッテ limv'匝可#0 .(normガ規 l z n l l= llyxny ー 111~IIYII l l x n l lI I Y 1 1 1ナル故.) 準化サレテアレバ l

w

§3. 前節ノ結果ヲ用ヒテノルム環ノ表現ニツイテ注意ヲ述ベテオキマス.但シココ デ表現トハ普通ノヤウニノルム環 R カラ有限次ノ行列ヘノ K 上ノ環トシテノ準同型 篤像ヲ云フモノトシマス.一般二表現ヲ

( 2 )

X

→ T(x)={切(エ)}豆=1 , 2 ,. . ., n

トシ,ソノ核即チ T (x)=0ナル如キ x ノックル両側 I d e a lヲ MT 卜書クコトニシ マス. { T ( x } }全体ハ行列環トシテ又ーツノノルム環卜考ヘラレマス.コノトキ ( 2 )ナル寓 像ガ連続ナラバ MTハ R デ閉ヂタ I d e a lトナリマスガ,コノ逆モ成立シマス.何トナ .R .H i l f s s a t z4 ノ方法デ R/MT~{T(エ)}二 norm I I X I I レバ MTガ閉ヂテヰレバ N ヲ入レレバコノ norm二封シテハ ( 2 }ハ明カニ連続篤像トナリマス.(*編者註 は xの属する剰余類をあらわす記号である.)然ルニ定理 1ニヨレバ { T ( x } }ノ normハ スベテ等値デスカラ I I 叫カラ I I T ( x ) I I *ヘノ窟像モ連続トナリマス.即チ適常二常数 C>Oヲトレバ 鼻

.x

( 3 )

llxll~!IXII~

1 — IIT(x)II*. C

48

[ 5 ]

特二

( 4 )

Cllxll~lt,;(x)I.

マトメテ云ヘバ 定理 3 .ノルム環 R ノ表現 ( 2 )ガ連続ナルタメニ必要且ツ十分ナル條件ハ核 MTガ R デ閉ヂテヰルコトデアル.コノトキ R ノ任意ノ x二対シ適営二常数 C>Oヲ卜 4 )ガ成立スル. レバ (

§4. 一般のノルム環デハトテモ手ガ出ナイノデ以下我々ハ常二次ノ如キ俵定ヲオイ テ進ムコトニシマス.

t , l rヲ R ノ任意ノ maximal左 乃 至 右I d e a l トスレバ R / I t , R / I rハ常 偲定 A.l 二有限次元デアル. サウスレバ先ヅ

.俵定 Aガ満足サレテヰレバ R ノ任意ノ maximal両側I d e a lI二封シ R/I 定理 4 モ亦常に有限次元,従ッテソレハ K 上ノ全行列環二同型デアル.

( 5 )

R/l~Kn ・

証明.Iヲ含ム m aximal左, . I d e a lヲ I 、トセヨ.俵定ニヨリ R/L 、ハ有限次元デソ レハ R ノ元ヲ左O p e r a t o r トシテ有スルカラ R ノ表現加群卜考ヘラレマス.コノ表 現デ 0二対應スル両側 I d e a lヲ I ' トスレバ明カニ I 'っIデスガ Iハ maximalナル ' .R/1=R/1'ガ有限次元ナルコトハコレカラワカリマスガソレ)ヽ K 上ノ 故 I=I 箪純環デスカラ R/I~Kn ・ コノ証明カラ又直チニ次ノコトガワカリマス. 定理 5 .俣定 Aノモトニ任意ノ maximal左ーI d e a ll eハアル maximal両側I d e a lI ヲ含ム. m aximal右I d e a lI r二関シテモ同様. 注意 .R二於テ定理 5ガ成立シテヰテモ必ズシモ逆二 Aガ満足サレテハヰマセン. 、ヵ I r (サウ云フ実例ヲツクルコトガ出来マス)又定理 5ヲ云フニハ俵定 Aノウチ I カドチラカダケガ成立シテヰレバ十分デスガ,ソレラノ條件ガ果シテ独立ナモノカド ウカハヨクワカリマセン. 定理 6 . 俵 定 Aノモトニ R ノ元ェノ左逆元ハ又ェノ右逆元デアリ逆モ成立スル. ョッテコノ唯一ノ逆元ヲ x-1 トカクコトガ出来ル. 証明• X ガ左(右)ー逆元ヲ有スルタメニ必要且ツ十分ナル條件ハェガ任意ノ m aximal 左(右)ーI d e a l二含マレヌコトデアリマス.( N .R .S a t z6 ) 定理 5ニヨレバソレハ又任 意ノ m aximal両側 I d e a lIヲトルトキ”ガ R/1二於テ maximal左(右)ーI d e a l . : : . 含マレヌコトト同ジデスガ R /l!=:Knデハソレハ又 maximal右(左)ーI d e a l二含マ レヌコトト等値(即チ Kn二於テ行列式ガ 0デナク逆元ヲ有スルコト)デスカラコレ

4 9

[ 5 ]

V匝可-→



サテ可換ノ場合卜同様二 0( n o o )ナル R ノ元 xノコトヲ一般零巾 元卜呼ブコトニスレバ次ノ定理ガ成立シマス.6 ) 定理 7 . 俣 定 Aノモトニ xガ一般零巾元ナルタメニ必要且ツ十分ナル條件ハ R ノ 任意ノ m aximal両側 I d e a lJヲトッタトキェガ R/1二於テ普通ノ意味デ零巾トナ ルコトデアル.

aximal両側 I d e a lIヲトッタトキ ( 5 )ニヨル R/1ノ既約表現ノーツヲ 証明. m T 1 ( x ) トカクコトニシマス. maximalI d e a lハ閉ヂテヰマスカラ定理 3ニ ョ リ 乃 ( x ) ハ連続表現デス. サテ任意ノ xER二封シ

1 1 x n 1 12 :I I T 1 ( x n ) I I= I I T 1 ( x r 1 1 ナル故 T 1 ( x )ガ零巾ナルコト,即チ x ガ R/1デ零巾ナルコトハ必要條件デアリマス. 1 ( x )ガ零巾デアレバ任意ノ入 E K二封シ巧(入x )=入乃(エ)モ亦零巾ナル故 逆二 T m ヲ十分大キクトレバ(入xre1. コレカラ eー泣ハ決シテ maximal左I d e a l ,l e 二含マレヌコトガ証明サレマス. 何トナレバ e-入x E I 1 . トスレバ.定理 2ニヨリ I tっJナル maximal両側ーI d e a lI ヲトリ(入xre1トシマス.サウスレバ(入x )m-l(e-入x )=(泣) m-1-(入 エ ) mE1 1 . カラ(入x )m-lEl eヲ得,同様ニシテ(入x ) m 2 ,・・・,(入x ) ,eEI t トナリ, コレハ不 合理.故二 e-入X ¢ _f 1 . . 従ッテ e —入ェハ左逆元ヲ有シマスガ定理 6 ニヨリ (e —入x)-1

ハ凡テノ入 E K二対シ存在スルコトニナリマス.コレカラハ可換環ノ湯合卜全ク同 e-泣)ー 1 ガ入ノ整函数ナルコトカラ展開 様デ (

(e-泣)ー l=e+泣 + 入2 丑+・・・ ニ於テ I I 入n x n l l→ 0 . ヨッテ

面n v ' 転可$肉・ 入ハ任意ダカラ

l i m祈戸『 =0 トナリマス.

aximal両側 I d e a lノ共通 I d e a l サテ再ピ可換環ノ場合卜同様二 R 二於ケル凡テノ m ヲ R ノ根基 ( R a d i k a l ) ト呼プコトニスレバ上ノ一般零巾元ノ全体ガ必ズシモ根基ト 一致スルワケデハアリマセンガ(例ヘバ R=Knノ場合ヲ見レバヨイ)次ノ定理ガ成 立シマス. 定理 8 .俵定 Aノモトニ R 二於ケル一般零巾元ノ全体ヲ N トスレバ根基 Noハ N 二含マレル最大ノ左(右) I d e a l .従ッテ最大ノ両側 I d e a lデアル. 証明.Noガ N 二含マレルコトハ前定理ニヨリ明カデスカラ N 二含マレル任意ノ I d e a lL 、ガ No二含マレルコトヲ云ヘバ十分デス.サテ任意ノ m江 i m a l両側 I d e a l 左Iヲトルトキ I ヽ・I/Iハ R/1ノ左ーI d e a lデスガ I ヽc Nナル俵定ニヨリソレハ零巾 元バカリカラ成ルモノデ,従ッテ I t ・ I=I ,7) I tcI .Iハ任意デスカラ[、 CNo. 定理 8ニヨリ特二恨定 Aノ満足サレテヰルノルム環デ一般零巾元ガ 0以外二存在シ ナイ場合ニハ No=O. 即チ Rハ準輩純環デアルコトガ云ヘルワケデス.

[ 5 ]

50

§5. 以上ノ如ク候定A ヲ承認スレバ,ノルム環ノ理論ハ代数的ナ部分二関スル限リ アル程度マデ可換ナ場合卜同様二進ムコトガ出来マスガ,位相的部分,即チ m a x i m a l I d e a l二 T o p o l o g i eヲ入レテ論ズルトイウヤウナ部分ハソノママデハ巧ク行キマセ ン.コレハ特ニ一般ノルム環ノ表現ノ問題卜関聯シテ今後二残サレタ問題デアルト思 ヒマス. 次二俵定 A二関シ少シバカリ注意ヲ述ベテ見マス.恨定 A ヲ満足シナイノルム環ガ 存在スルコトハ H i l b e r t空間二於ケル有界作用素カラ容易ニソノ例ヲツクルコトガ出 来マス.又俵定 Aヲ満足スル場合トシテハ次ノ例ガ知ラレテヰマス.

1 . R ガ可換ナル場合

2 .R / I oガ有限次元ナル如キーツノ両側 I d e a li I 。ガ存在シテ I 。二含マレル x二対シ y=x y ,A~ ・ y=yxナル O p e r a t o rA x ,A~ ガ完全連続ナル場合. テハ Aェ・ 1ハ G e l f a n dノ可換ノルム環ノ場合デアッテ, 2ハ深宮氏ガ c o m p a c t群ノ群環二於 テ述ベラレタ場合デアリマス.8 ) 一般ニドノヤウナ條件ガアレバ俣定 Aガ満足サレルカト云フコトハ難カシイ問題デ アラウト思ヒマス.

I I .群環 R(G) §1. 群 G ハ l o c a l l yc o m p a c t ,s e p a r a b l e トシ,μ ヲ G 上ノ右i n v a r i a n tナ H a a r ノ測度トシマス図 G 上 ノ 可 測 函 数 ニ ツ イ テ び ( p2 :1 ) ノ定義及ピソコニ於ケル

norm ( 6 )

{ f a1x(g)IPdg}

l l x ( g ) l l v=



ハ周知ノ通リ.

x ( g )EL 丸y ( g )E£1二封シ ( 7 )

j

xxy ( g )=z ( g )= x ( g h 1 ) y ( h ) d h G

トオケバ簡輩ナ計算ニヨリ 10)

{ 8 ) .

l l z ( g ) l l v : : : ;l l x ( g ) l l v ・ I I Y ( u ) l l 1

即チ z ( g )ハ賓際存在シテ LP二属スルコトガ知ラレマス. サテが卜 L P(p~1) トノ共通部分ヲ £(1,p) トシ x, yE £ ( l , p )二対シ 積ヲ定義スレバソレガ環ニナルコトハ容易ニワカリマス。ソコデ更ニ

( 9 )

( 7 )ニヨリ

紐 ( l l x ( g ) l l p ,l l x ( g ) j l i ) l l x l l= M

トオケバ { 8 )ニヨリ

{ 1 0 )

l l xxYI!~l!xlll!YI!

デ£ ( 1 , p )ハ鼠位を有スルト云フ條件ヲ除ケバノルム環ノ他ノ條件ヲ凡ベテ満シマス. ヨッテ £ ( 1 , p )二箪位 eヲ附加シテ

入e+x ( g ) , 入EK, x ( g )E £ ( l , p )

[ 5 ]

51

ナル全体ヲツクレバコレハ普通ノ様ニシテノルム環トナリマス.コレヲ R=R ( l , P ) ( G ) トカクコトニシマス .Gガ d i s c r e t eナ場合ニハ既二 £ ( 1 , p )二単位ガ存在シ £ ( 1 , p )自 身ノルム環トナルワケデスガ我々ハコレラノ場合ヲ匿別シナイコトニシマス. S e g a lハ R ( l , l ) ( G ) ヲ G ノ群環卜呼ンデヰルノデスガ R ( l , v l ( G )ハムシロ補助的 ナ意味ヲモツモノデ £ ( 1 , p ) ( G )ノ方ガ(有限群ノ場合ノ拡張卜考ヘテモ)群環ノ名ニ 相應シイノデハナイデセウカ.以下ノ我々ノ考察モ目標ハ £ ( 1 , P ) { G )ノ方二置イテ進 ミタイト思ヒマス. R < 1 , P > ( G )ノ中デ考ヘレバ £ ( 1 , P ) ( G )ハーツノ maximal両側 I d e a lMoヲツクリマ ス.又 x ( g )E £ ( l , p ) ,aEG 二対シ

( 1 1 )

a x ( g )z=x ( g a ) , x a ( g )=x ( a g )

ト定義スレバ a x,xaモ亦 £ ( 1 , p )二属シ特二 I l a 叫1=11叫. J

§2. 以下二於テハ £ ( 1 , P l ( G )ノ表現卜 G ノ表現トノ関係ヲ考ヘルコトニシマス.コ コニ表現トハ普通ノ意味ノ有限次行列ニヨル表現ヲ云フワケデスガ特二 G ノ表現トシ テ0 ー表現ヲモ許スコトニシマス.即チ G ノ任意ノ元 a , b二対シ D(a)D(b)=D ( a b ) ヲ満足スル行列系 { D ( a ) }ハスベテコレヲ G ノ表現卜呼ブコトニシ G ノ輩位ガ鼠位 行列二対應スルト云フコトハ要求シナイコトニシマス. サテ我々ノ目標ハ次ノ定理デアリマス. 定理 9 .£ ( 1 , P ) ( G )ノ連績表現 x ( g )→ T(x) 卜 G ノ有界連績表現 a→ D(a) トハ 次ノ如キ意味デー封ーニ対應スル. i )L < 1 , P > ( G )ノ連績表現 x ( g )→ T ( x )ガ輿ヘラレタトキ適常二 G ノ有界連績表現 a→ D(a) ヲトレバ任意ノ x ( g )E £ ( l , p )二対シ

( 1 2 )

jx(g)D(g)dg

T ( x )=

G

トナル.コノ様ナ D (a)ハ T ( x )ニヨリー意的二定マリ,コレヲ DT(a) トカクコト ニスル. i i ) G ノ有界連績表現 a→ D(a)ガ輿ヘラレタトキ ( 1 2 )ニヨリ T(x) ヲツクレバ x ( g )→ T(x)ハ £ ( 1 , P ) ( G )ノ連績表現ヲ興ヘル. D ( a )二封應スルコノ表現ヲ T v ( x ) トカクコトニスル. i i i )i ) ,i i )ノ封應ハ互二他ノ逆ニナッテヰル.即チ任意ノ T ( x ) ,D(a)二封シ

TvT=T, D⑬ =D. 2 ( x ) i v ) 上ノ封應デ等値ナ表現ハ等値ナ表現二封應スル.即チ AT1(x)A-1= T ナラバ ADT1( a ) A 1= D叫 a ) ,又 BD1(a)Bー1= D 2 ( a )ナラバ BTv1( x ) B 1= T v 2 ( x ) . 証明.各段二分ケテ証明シマスガ面倒ナノハ i )ダケデアトハコレカラ直チニ出マス. i ) £ ( 1 , v l( G ) ノ表現ヲ x(g)→ T ( x )= {切 ( x ( g ) ) } トオイテ切 ( x ( g ) ) ヲ考ヘテ 見マス.

T ( x )ハ連績表現ナル故定理 3ニヨリ l t i 3 ( x ( g ) ) I: 5C l l x ( g ) I Iナル常数 C ガ存在シ マス.特二有限ナ測度ヲ有スル場合 E ノ 特 性 函 数 咋 ( g )ヲトレバ

( 1 3 )

l t i j ( X E ( g ) ) Iさ C l l x E ( Y ) I I=Cmax(μ(E),μ(E)り

[ 5 ]

52 ヨッテ適嘗二可測函数 U i j ( g )ヲトレバ 11) 上ノ如キ任意ノ

l

=

{ 1 4 )

t i ; ( x E ( g ) )

E 二封シ

=j 咋 ( g )叩 ( g ) d g . G .

U i j ( g ) d g

切( x ( g ) )ガ連績ナルコトヲ用ヒレバコレカラ一般ニ

j

如 ( g ) )= x ( g ) u i i ( g ) d g . G

=

即チ U ( g ) { U i j ( g ) } トオケバ

j

T ( x )= x ( g ) U ( g ) d g .

{ 1 5 )

G

T ( x )ハ表現デスカラ £ ( 1 , P ) ( G )二含マレル任意ノ x ( g ) ,y ( g )二封シテ T ( x ) T ( y )= T(xxy ) ,即チ

jx(g)U(g)dgj y ( h ) U ( h ) d h=j x ( g h 1 ) y ( h ) U ( g ) d g d h . ヨッテ

J

J

痘) y(h)U(g)U(h)dgdh= x ( g ) y ( h ) U ( g h ) d g d h .

有限ノ測度ヲ有スル集合ノ特性函数ハ凡テ £ ( 1 , P ) ( a )二含マレマスカラ上式カラ ( g ,h ) -

測度 0 ヲ除キ U ( g ) U ( h )=U ( g h )

( 1 6 )

ヲ得マス.ヨッテ適営二測度 0ナル集合 Xoヲトレバ h

( G )ノ両側 I d e a lデ £ ( 1 , P ) / Mハ有限次元ナルモノトスル.コ 定理 1 ノトキ M ガ £ ( 1 , P } ( G )デ閉ヂテヰレバ £ ( 1 , P ) / Mハ準輩純環デアリ,逆二 £ ( 1 , P ) / M ガ準輩純ナラバ M ハ £ ( 1 , v > ( G )デ閉ヂテヰル. コレラハイヅレモ有限群ノ群環二関スル定理ノ拡張卜考ヘラレマス.

[ 5 ]

56

サテ £ ( 1 , P ) ( G )ノ既約表現ノ各類 { T ( x ) }ニハソレゾレ一定ノ maximal両側 I d e a l T ( x ) }ハ又定理 9ニヨリ G ノ有界既約表現類 { D ( a ) } トー封 Mrガ卸應シマスガ { 一封應シマスカラ次ノ定理ヲ得マス. 定理 1 3 .{D}ヲ G ノ有界可測既約表現ノーツノ類, D(a) ヲソレニ属スルーツノ 既約表現トスルトキ

J

x ( g ) D ( g ) d g=0

G

ナル如キ x ( g )全体ハ £ ( 1 , P l ( G )二於テ {D}ニヨッテ一定スル maximal両側 I d e a l M{D} ヲックル. {D}→ M{D}ナルコノ封應ニヨリ G ノスベテノ有界可測既約表 現ノ類 {D} 卜 £ ( 1 , p )/Mガ有限次元トナル如キ £ ( 1 , p )( G ) ノスベテノ maximal両 側 Ideal トハー封—酎應スル.

一般二 G ノニッノ有界可測表現 D 1 ( a ) ,D 2 ( a ) ガ輿ヘラレタ場合,ソレラニ含マ 1 ( a )二含マレル既約成 レル 0ナラザル既約部分が全体トシテ一致スルトキ(即チ D 分ハ凡テ D 2(a)二含マレ,逆モ亦成立スルトキ) D 1 ( a ) 卜D 2 ( a ) トヲ同ジ類二入レ ルコトニスレバ,コノ様ニシテ G ノ凡テノ有界可測表現ヲ類別スルコトガ出来マス ガ,コノヤウナ表現類トソレニ封應スル £ ( 1 , P l ( G )ノ表現ノ核トナル £ ( 1 , P l ( G )ノ両 側I d e a l トガー封ーニ封應スルコトハ明カデス.貰際 D(a)二含マレル既約表現(類) ヲD 1 ( a ) ,. . ., D k ( a ) トスレバ a→ D(a)二封應スル £ ( 1 , P l ( G )ノ表現 x→ T瓜x ) ノ核ハ M{D か...,M 図}ノ共通 I d e a l トナリマス. £ ( 1 , P ) ,p=l ,2 ,. . .ニ於イテ p ヲ変ヘテモ G ノ表現類ハ一定デスカラソノ封應ノ サセ方カラ次ノ定理ヲ得マス. 定理 1 4 .£ { 1 , 1 > ( G )=が ( G )二於テ L1/Mガ有限次元トナル如キ任意ノ閉ヂタ両 d e a lM ヲトリ 側 I Mn£(1,P>(G)= Mp, p= 1 ,2 ,. . . トオケバ Mpハ £ ( 1 , P ) ( G )二於テ閉ヂタ両側 I d e a lデ , £ ( 1 , p )/Mpハ有限次元デアル. カクノ如クニシテ £ ( 1 , P l / M pガ有限次元トナル如キ £ ( 1 , P l ( G )ノ凡テノ閉両側 I d e a l ハp 1 , 2 ,…二射シ互ニー封ーニ酎應スル.コレラハ又 G ノスベテノ有限可測表 現の類 { D} ト一酎ーニ封應スル. ( 1 , P ) ( G )或ハ Gノ表現ヲ問題ニシテヰル限リ £ ( l , p )( p=1 ,2 ,…)ノド ョッテ £ レヲトッテモ同ジコトデ,ソノ場合二應ジテ便利ナモノヲ選ベバヨイワケデス.

=

§3. 次二始メニ述ベマシタノルム環 R < 1 , P ) ( G )即チ 入e+x ( g ) , x ( g )EL ( l , p ) ( G )

ノ様子ヲ調ベテ見ルコトニシマス.ソレニハ前節ノ結果ヲ用ビルコトニスレバ次ノ補 助定理ダケデ十分デス. 補助定理 2 .M ハ £ ( 1 , P ) ( G ) ノ両側 I d e a l ,£ C l , p )/ M ハ輩位元ヲ有シ且有限次元 ( l , P ) ( G )ノ両側 I d e a lM'ヲトレバ トスル.然ラバ適常二 R

( 2 8 ) ( 2 9 )

M

=M'nL(G)

R ( t , P ) ( G )=M'+L ( t , p ) ( G )

( 5 ]

57

トナル.而モコノ様ナ M'ハ唯一ツ定マル.

( 1 , P ) ( G ) ノ両側 I d e a lデスガ,掛ヶ算ノ定義ノ仕方カラソレハ又 証明 . Mハ £ R ( l , P ) ( G ) ノ両側 I d e a lデモアリマス.仮定ニヨリ £ ( 1 , p )/Mハ訊位元ヲ有シマスカ ラソレニ封應スル £ ( 1 , P l ( G ) ノ函数ヲ x o ( g ) トシ, Yo= e-x o ( g ) トオケバ容易 ニワカルヤウニ R < 1 , P >/Mハ £ ( 1 , p ) / M 卜 {y , 。M}/M トノ直和トナリマス.ョッテ {y , 。M}=M' トオケバ M'ハ R(l,P)(G)ノ両側 Idealデ (28), (29)ヲ満足スルコ トハ明カデス. M'ノー意性ハ R ( l , p }/Mノ分解ノー意性ニヨリ明カデス. サテ上ノ M'ヲトレバ ( 2 8 ) ,( 2 9 )カラ R ( l , p )/M'~ £ ( l , p }/M トナリマスカラ特二 M ガ£ ( 1 , p }デ maximalデアレバ M'ハ R < t , p )デ maximal トナリマス. R ( l , P ) ( G ) ハノルム環デスカラソコデノ m aximalI d e a lM'ハ閉ヂテヰマス.ヨッテ ( 2 8 )ニョ ( l , P ) ( G ) ,従ッテ £ ( 1 , p ) ( G )デ閉ヂタ I d e a l トナリマス.(前節参照) リM モ R ( 1 , P ) ( G )ノ閉ヂタ両側 I d e a lM15l 卜R < 1 , P l ( G ) ノ Mo= 又コノ補助定理ニヨリ £ £ ( 1 , P l ( G )二含マレナイ閉ヂタ両側 I d e a l トガ ( 2 8 ) ,( 2 9 )ノ関係ニヨリ互ニー射ーニ 封應スルコトモワカリマス.コレラノコトカラ前節ノ結果ヲ用ヒレバ容易二次ノ諸定 理ヲ導クコトガ出来マス.

5 .R ( l , P ) ( G ) ノ任意ノ連続表現 y→ T ' ( y )二酎シテハ G ノ連続有限表現 定理 1 a→ D(a)ガー意的二定マッテ y =入e+x ( g )ナルトキ ( 3 0 )

J

T ' ( y )=T ' (入e+x(g))=入D ( e o )+

x ( g ) D ( g ) d g 1 6 )

G

トナル.逆二 G ノ任意ノ連続有界表現 a→ D (a)二射シ ( 3 0 )ニヨリ T ' ( y ) ヲック ' ( y )ハ R < 1 , P > ( G )ノ連続表現ヲ輿ヘル.カクシテ表現 T ' ( y ) 卜 D(a) ト レバ y→ T ハー封ーニ封應スル.

6 .R ( t , p ) ( G )ノ連続表現ハ完全可約デアリ逆二完全可約ナ R ( l , P ) ( G )ノ表現 定理 1 は凡テ連続デアル. 7 . M1ハ R < 1 , v l ( G ) ノ両側 I d e a lデ R < 1 , v >/M1ハ有限次元ナモノトス)レ. 定理 1 コノトキ払ガ R < 1 , P > ( G )デ閉ヂテヰレバ R ( l , p }/M1ハ準凱純環デアリ,逆モ亦成 立スル. 定理 1 8 .{D} ヲ G ノ有界可測既約表現ノーツノ類, D(a) ヲソレニ属スル任意ノ 既約表現トスルトキ

(y,D)=

J

平+

x ( g ) D ( g ) d g=0 ( nハD(a)ノ次数)

G

ナル如キ y =入e +x(g)ノ全体ハ R ( l , P } ( G )二於テ {D}ニョッテ一定スル maximal 両側 I d e a lM{D}ヲツクル.コノ封應ニヨリ G ノ凡テノ有界可測既約表現ノ類 {D} トR ( l , p )/M1 ガ有限次元デ M1-#Moナル如キ R ( l , P ) ( G ) ノ凡テノ maximal両側 I d e a l トハー封ーニ封應スル. コレガ S e g a lノ論文二於ケル定理 3デス.1 7 ) 上記ノ定理ハ S e g a l ト同ジャゥニ云 ヒ表ハシタノデスガ.モシモ前ノ様二 O ー表現ヲモ既約表現ノ仲間二入レレバ定理ノ 後半二於ケル M1-#Moナル條件ハ取除クコトガ出来マス. 同様ニシテ定理 1 4二相営スル定理モ得ラレマス.

58

[ 5 ]

以上ノ如ク我々ハ前節ノ定理 9-1 3( 1 4 ) ヲ用ヒテ定理 1 5-1 8ヲ証明シタノデ スガ,コレラノ諸定理ハ勿論直接ニモ証明サレマス.証明ノ方法ハ前節卜全ク同様デ, シカモコノ際 R < 1 , P > ( G )ニハ輩位元ガ存在シマスカラ £ ( 1 , P ) ( G )ノ場合ヨリムシロ簡 輩二出来マス. 5-1 8カラ前節ノ結果ヲ導カウトスルト結局前節二於テシタト同様 然シ逆二定理 1 ( l , P ) ( G ) ナ考察ヲ繰返スコトニナル様二思ハレマス.ソシテ定理ノ本質的ナ部分ハ R ヨリ £ ( 1 , P } ( G )ニアルト思ハレマスノデ £ ( 1 , P } ( G )ニツイテ稚々面倒ナ証明ヲシタワ ケデス.

§4. 定理 9ノ應用トシテび ( G )二属スル表現ヲ考察シテ見マス.ココニ表現 D(a)= 的 ( a ) }ガ び ( G )二属ストハ各函数 d i j ( a )ガ び ( G )二属スコトヲ略稲シタノデス. 先ヅ D (a)ガ じ ( G )二属スル表現デアレバ

苫 如 (g)如 (h)jdgdh

Lxal d i j ( g h ) j d g d h= L x aI

三 吝f 山( g )如 ( h ) l d g d h予 = j,如 (g)ldg.J l d k j ( h ) j d h

く 00.

一方

L Gロg:::~~~!(::l~:;W)ldgdh I d , ;

故二 μ (G)= o o ト俵定スレバ fald 匂; ( g ) l d g= o .即チ測度 0 ヲ除イテ dij(g)=o . i , jハ任意デスカラ適常二 g=90 ヲトレバ凡テノ i , j二関シ,d i j ( g o )=0 , 即チ D ( g o )=0 . ョッテ任意ノ gEG二射シ D (g)=D(gg0り D ( g o )=O . 即チ D(a)ハ

0表現デナケレバナリマセン. D(a) ガび (p~2) 二属スル場合モ Holder ノ不等式

苓 心 0)如 (h)I'2ナラバ差嘗リ適嘗二 μ ( E ) : : ;2ナル E ガ存在シテ E ——• L ノ外デハ U i j ( g )ハ有界 .E上デハ i i ; 1 )二属スルコトガワカル.実ハコノト

( !+¼=

キモ U i j ( g )ガ有界ナルコトハ以下ノ証明二示ス通リデス. 1 2 ) 後の應用ノタメニ我々ハ D ( a ) ,D ' ( a )ガ可測デアルコトダケヲ俣定シテオキマス. 1 3 ) ココデコノコトヲ用ヒルノハマヅイヤリ方デアッテ我々ノ立場トシテハ逆二 T(x) ガ完全 可約ナルコトカラ D (a) ノ完全可約性ヲ出シタイ様二思ヒマス.ソシテコレハ特二 G ガ compactナル場合ニハ比較的簡輩二証明サレマス.コレラノコトニ関シテハ又別二述ベテ 見タイト思ヒマス. 1 4 ) 一般二連績デナイ表現ガ存在スルモノナノカドウカヨクワカリマセン. 1 5 ) M ガ閉ヂテヰレバ定理 1 2ニヨリ i < 1 , P l/Mハ準箪純,従ッテ輩位ガ存在スルコトニ注意 1 6 ) 群 G ノ輩位ヲ R < 1 , P l ( G )ノ躁位卜匿別シテ e o トカキマシタ. 1 7 )S e g a lハソコデ G ガ compact或ハ a b e l i a nナル場合ダケヲ云ッテヰルノデスガ,ソレハ 恐ラク R(1,1J(G) ノ庄乞~maximal i d e a lヲ問題ニシテヰルタメデアリマセウ. 1 8 ) G ハ compactナル故 μ(E)=μ(Eー1 ) トナル. 1 9 ) 註4 ) ノ論文二於ケル定理 5参照. 2 0 )S e g a lノ定理 1 . 又深宮氏ノ論文参照.ソコデ直接簡草ナ証明ヲサレテヰマス. 2 1 )註 4 )ノ論文二於ケル定理 6 .

[6]

On the structure of infinite M -groups Japanese J. Math., 18 (1943), 709-728.

(Received February 21st, 1942.)

§ 1. Introduction.

In a previous publication°!, I have studied the structure of finite M-groups. The purpose of the present paper is to extend the results of that paper to the case of infinite groups. A group @ is called M-group when the lattice £(@) formed by all subgroups of @ is a modular lattice< 21 : it holds namely for arbitrary subgroups \ll, ?8, [ with l!l c:: [, of @ the modularity equation (1)

In this case, if we make each subgroup x lying between ~( v ~ and 't\ correspond to the subgroup xr-.~{. and each subgroup ~)) lying between '!l and \II,.-,.18· to the subgroup vv?B, we have a lattice isomorphism between quotient lattices 9Iv$/58 and '.11/\ll,.-,.~: ( 2)

It follows then, that the length of any principal chain which connects two subgroups of @ is always equal to, each other(3). A group is called "group of finite length ", when all its principal chains which connect the whole group with the identity, have always the same and finite length. For example any finite M-group is a group of finite length. But it is not yet decided whether a group of finite length is finite or not. This is a special case of the unproved assumption, that a group is finite when its subgroups satisfy the both chain conditions1-41 , The structure of a finite M-group is given by the following theorems< 51 : K. Iwasawa. Ober die endlichen Gruppen und·die Verbande ihrer Untergruppen, Journal of the faculty of science, Tokyo Imperial University, I, vol. IV, part 3. (1941)referred to as G.V. (2J See G. V. p. 171. For the general .lattice theory see G. Birkhoff, Lattice theory (1940). This assumption is indeed valid for those groups which can be isomorphically represented with matrices on some commutative field. Cf. I. Schur, Sitzungsber. Preuss. Acad. Wiss. 1911, p. 619-627. (5) See G. V. Satz 2, 3, 18.

61

62

[6]

710

Kenkiti

IWASAWA

>

1. A p-group (p 2) ~ is an M-group if and only if it has the following structure : ( i ) ~ contains an abelian normal subgroup ~r, (ii) ~/~l is cyclical, (iii) for a suitable generator T of ~/~t we have for any element A of~ ( 3)

where s means an integer which is uniquely determined by ~ and ?I, and is independent of A. 2. A 2-group is an M-group, if and only if it has either the same structure as in 1 with an additional condition s ~ 2, or it is a Hamiltonean 2-group, i.e. it is the direct product of a quaternion-group and an abelian group of type (2, 2 ... , 2). 3. Let p, q be two different prime numbers, p q. A group @ of order p" q~, which is not the direct product of their Sylowgroups, is an M-group if and only if it has the following structure : ( i ) A p-Sylowgroup ~ is normal in @ and is an abelian group of type (p, p ... , p). (ii) A q-Sylowgroup @ is cyclic; @ = {Q}. (iii) For an arbitrary element P of ~ it holds

>

( 4)

QPQ-'

= pr

where r· is a definite integ~r which depends on1y on the group satisfies following conditions (5)

nf= 1 mod. p,

@

and

rq == 1 mod. p.

4. A finite group @ is an M-group if and only if it is the direct product of groups of mutually prime orders which have structures mentioned above in 1-3. Now let @ be an arbitrary M-group and A and .J3 its elements of finite orders. We prove first that AB is also of finite order. Let ·~ = {A}, ~ = {B}. From (2) it follows easily that ~v~ is of finite length; consequently it contains no element of infinite orde,r, because· an element C of infinite order would give rise to an infinite seqpence of subgroups {C}, {C2}, {O'}, . . . . As an element of \llv~ AB is of finite order. Thus we obtain the following Theol'em 1. In an M-group @, elements of finite orders form together a characteristic subgroup ~ of @. rt,/~ is then an M-group whose all elements except the identity are of infinite orders.

63

[6]

711

On the structure of i11finite M-gi·ou7Js.

+

In the following we treat two cases separately, @ = Cf, and @ fr . In § 2 we study the case @ = G:. On that occasion we must employ the assumption mentioned above, i.e. that a group of finite length is finite. Therefore the theorems which are to be obtained in § 2 are available only for those groups whose subgroups of finite length have finite orders, for example locally finite groups, quasi-Hamiltonean groups etc. But if that assumption could have been proved anyhow, then our results would give the structure of general M-groups whose all elements are of finite orders. In § 3, we determine the structure of an M-group which has at least one element of infinite order. It is to be noted that in this case @+ G:, we can carry out our investigation without any assumption. This may deserve some interest. In fact we prove a special case of the above assumption and by help of this lemma we can avoid the relating difficulties after some delicate considerations. § 2.

M•groups without element of infinite order.

1. In this paragraph we always consider such an M-group @ that its elements are all of finite orders and its subgroups of finite length are finite. We have then at once. Lemma 1. If '2( and 5B are finite subgroups of @, then \1Ivi1 is also finite. Finite subgroups of @ thus form a sublattice of L(@). Proof. From (2) it follows easily that '2( v$8 is of finite length, and so is finite by .the assumption< 6 >. Now let p be a-:iy prime number. We assume first that any element of p-power order is commutative with all elements which have orders prime to p. Let A, B be any element of order p", p~ respectively. According to Lemma 1 {A, B} is a finite group and by the assumption its p-Sylowgroup is normal. While A, B are contained in this p-Sylowgroup, ·{A, B} itself is a p-group and AB has also a p-power order. All elements of p-power orders form therefore a characteristic subgroup of @ : the p-component lJ,3 of @, We assume next that an element P of order pm is not commutative with an element Q of different prime power order q.,._ As before it is easy to see that {P, Q} is a finite group of type 3 in § 1. We may suppose for instance, exchanging P and Q if necessary, that a p-Sylowgroup ~o is an abelian normal subgroup of type (p, p, ... , p), and a q-Sylowgroup is cyclic. Now let P' be another element of p-power The converse is also true; that is to say, this- lemma is equivalent with the assumption mentioned above.

712

order.

Kenkiti

IWASAWA

{P, P', Q} is also a group of type 3 in § 1. It follows therefore

PP' =P'P,

PP= P'P = 1,

that is, all elements of @ of p-power orders form an abelian characteristic subgroup ~. any of whose elements satisfy the relation

XP = 1.

( 6)

Then it is easy to see that for an arbitrary element Po of $ it holds ( 7)

QPoQ-:1 =Po,

r $ 1, mod. p,

r'l == 1, mod. p.

If we take another Q' of q-power order then {P, Q, Q'} is also of type 3 in § 1. We denote the normal p-Sylowgroup of this group by ~ 1 and a cyclic q-Sylowgroup by 0 1 = {Q*}. It follows then

( 8)

Q*PiQ*-1

= P[i,

r1

1'1== 1, mod.

$1, mod. p,

for all Pi in $, . If we have mod.

Q*u== Q

p,

ti,

it follows that 11,

pr1

=

Q*u PQ*-u

=

QPQ-1

=

pr ,

consequently (cf. (7), (8)) rf

EE

r , mod. p,

u$0, mod, q.

Q generates therefore also a q-Sylowgroup and it is

{P, Q, Q'}

= {ll31, Q}.

Any element of q-power order is thus contained in the subgroup {~, Q}. {~, Q} is therefore a characteristic subgroup of @. From above considera.tions we obtain the following theorem which corresponds to 4 in § 1. Theorem 2 •. An M-group @ which satisfies the assumption stated at the beginning of this paragraph is the direct product of following groups: ( a) .The M-group in which any element has an order which is ' a power of a fixed prime number. (#) The group {~, Q}, where ~ is an abelian group whose elements satisfy the relation

X"=l, and Q is an element of q-power order. p, q means then two different prime numbers. Between Q and an arbitrary element P of ~ it holds

65

[6]

713

Ou the strnct1l/'t! ul infinite 1'111-11 rou11;;.

QPQ- 1 = pr,

r $1, mod. p,

r'1

= 1,

mod.

P,

where r is an integer depending only upon {$, Q}. The orders of elements contained in different direct factors must be prime to each other. Conversely any group, which has the structure mentioned above, is an M-group which satisfies our assumption. The last part of the theorem is easily seen from 3, 4 in § 1. 2. According to Theorem 2 it suffices to investigate only such M-groups any of whose elements has an order which is a power of a fixed prime number. We first consider a 2-group @ which contains a quaternion-group 0. Let A, B be any two elements of the group and put GJ1 = {O, A, B}. According to 2 in §1, @1 is a Hamiltonean 2-group and it holds

BAB-1 = A" 1 The whole group @ is therefore also Hamiltonean and is the direct product of Q and an abelian group of exponent 2. Conversely any such group is obviously an M-group. Now let @ be a modular p-group which is neither abelian nor Hamiltonean. By the assumption @ ~ontains a finite non-abelian subgroup @1. We consider any finite subgroup @..i of @ containing a finite subgroup @~il, which has no abelian normal subgroup 9l~il with conditions

@i•> =

{~~il,

TJ•>,

So},

~!il

>

~~il.

Put

It follows then ~

= {~rt, Ts*, so} ,

and ~ is therefore one of ~!·> in (10): written in the form

~;

= ~l!io>.

Now @~io> can be

with 2{~;o).

= ~ r,@~iol ~ 2{J' r, ~· = 2{1i

which contradicts our assumption. We have thus found a finite subgroup perties: (a) @* is expressed in the form @*

= {~{*,

= @*

with following pro-

T, so} •

Any finite subgroup @s, which contains @*, can be expressed in the form

(,8)

}, ttt rn 80 '216= {S>l'. u.a,.L6,

(ry) As @s

@5

0) ,

o(@*)

,;i1 ar*, U 6 --. --«

ttt

\216

= .-i1vttt* «6 \!Y

= pk = Max. 0(@4) .

= 'llsv@* =

{9!6 , T} we can put moreover in (,8) T 6 = T.

[&]

68 716"

Kenkiti IWASAWA

Now let @.,, @s be any finite subgroups such that @s::::, .@.,::::, @* • ·@.,

= {~7 ,

T, Bo} , ~., ~ ~*

@a

= {'2ls ,

T, Bo}.

~s > ~* •

Put ~7v~s

= {'2ls,

T s>tt }

,

'f P"' = A..As, A7 e ~7-, Ase We.•

Take an element Ao of order pk from 121*. As 121* is contained in 1ll'.7 " 1218 , we have Ao(l+p"O)P"

= TP"'AoT-"' =• A7AsAoAii1Ai"1 == Ao'

and consequently (11)

As the order of any element 121s is at most pk by the assumption, it is easy ,to see from (11) tliat TP"' is commutative with any element contained in lll8 • 121.,vl2ls is therefore an abelian group and we can write @a=

{2£.zv'2£s, T, Bo}.

If we take @., = @a, above consideration shows that there is unique maximal abelian normal subgroup 21s such that @s

= {~*,

I', Bo},

~::::, ~*.

In the following if any.@9:::, @* is expressed in the ·form @s

= {~,

T~ Bo} ,

we take then alawys as 2le the maximal abelian normal subgroup con.: taining 121*. For any .finite subgroup& @9 , @1e with @10 :::,

@s

= {~9,

@g ::::, @* ,

T, so} ,

@10 =-

{~Jo, T, Bo},

it is then easy to see from above consideration, that (12)

Now for arbitrary element.a X; .Y in @. put

[6]

69 On the structure of infinite M-groups.

= {@*, @y = {@*,

X}

@x

@x. y

=

717

{%ex, T, So}

Y} == {%Cy, T, so}

= {@*,

X, Y}

= {2lx, Y,

T, So}.

It follows from (12) that ~x

= %ex, yr"\@x,

AE 12lx, y is abelian, elements of 12lx and \lly are commutative with one another. The group

is accordingly an abelian group, and for any element A in \It we have TAT- 1 = A 1+p80 •

It is also clear that @= {~. T}.

We have thus proved the first half of the following theorem, which is an extension of results 1, 2 of §1 in the case of infinite groups. Theorem 3. A non-abelian p-group< 8 > @, whose subgroups of finite length are all finite, is an M-group, if and only if it has the following structure. (a) p>2. ( i) @ contains an abelian normal subgroup \ll.

The order of elements of l2l is bounded. (ii) @/'el. is a cyclical group of order pm. (iii) For a suitable generator T of @/\ll we have for any element A in~

where s means an integer which is uniquely determined by @ and \II and satisfies the inequality

s+m2n, (iv)

if the maximum of the. order of elements in \ll is p". Tm = To is an element of ~ and it holds p•

To = 1. l 8 l A group is called p-group generally when the orders of its elements are all p-powers.

70

[6]

718

(/:1)

Kenkiti

p

IWASAWA

= 2.

is either a Hamiltonean group, or a group which has the same structure as mentioned in (a) with an additional condition s~2. Corollary. A modular p-group, y1hose subgroups of finite length are finite, is either abelian or meta-abelian. Proof. We have only to prove the second half. It is easy to see that in any group @, which has the structure mentioned above, subgroups of finite length are all finite. Let A, B be arty element of @. The finite group {A, B} is then of type 1 (or 2) in !§1. There exists consequently such integers x, y that @

BA== A"'B".

is therefore quasi-Hamiltonean< 9 >, accordingly modular a fortiori. In a quasi-Hamiltonean group any subgroup of finite length is, as readily to be seen, finite. If we remark therefore that a finite quasiHamiltonean group is the direct product of their Sylowgroups and the groups given in Theorem 3 are all quasi-Hamiltonean, we have immediately the following Theorem 4. A group, whose elements have. all finite orders, is quasi-Hamiltonean, if and only if it is the direct product of 'an abelian group and such groups, which have the structure stated in Theorem 3, where the orders of elements of different direct factors are prime to each other. A quasi-Hamiltonean group is therefore always either abelian or meta-abelian. @

§ 3.

M-groups which contain elements of infinite orders.

1. We now determine the structure of M~groups, which contain elements of infinite orders. This time we can carry out our investigation without any assumption. Lemma 2. Let A, B be any two elements of an M-group @, of which A has an infinite order. If {A}/'"\ {B} = 1, then

(13)

{A, B}, {A2, B} , {A3, B}, ... , {A'\ B} , ..• , {B}

are different from one another and e:x:haust all subgroups of @ which lie between {A, B} and {B}. {9)

A group

@

is called "quasi-Hamiltonean ", if for any subgroups \ll, l8 we have \lll8

= !!Ill! = lltv!!l .

For the relation between an M-group and a quasi-Hamiltonean group see G. V. i 3.

71

(6)

719

On the. st·ructure of in.finite M~g1·oups.

Proof. Put~= {A}, ~ = {B} and apply (2). As {A}""{B} the subgroups of {A} are given by

{A}, {A2 }, {A 3},

••• ,

=1

and

{A"}, ... , {1},

we can immediately obtain. the lemma. By help of this we prove the following Lemma 3. Let A be any element of infinite order and B any element of finite order of an M-group @. It holds then in taking suitably an integer r

ABA- 1 = B", (r, m)

(14)

= 1,

where m is the order of B. Proof. The group {B, ABA- 1} lies between {A, B} and {B} and is therefore one of the subgroups (13). But as it is generated by elements of finite orders, it contains no element of infinite order (cf. Theorem 1). It must be then {B, ABA- 1} = {B} and we can readily obtain (14). Lemma 4. Let A, B be any two elements of an M-group @, of which A has an infinite order. If it holds for suitable integers a, fJ

BA"B- 1 =A',

(15)

then it must be a= /3 and A" commutes. with B. Proof. If {A},-..{B} == {AT}=l=l, then it follows from (15)

whence

o'Y ;=

fJy, a

= fJ.

Let {A} rs. {B} {A'", B}

=

= 1.

It is

{A 0, B}

and from Lemma 2 we conclude a= ±f). Assume a=-fJ. From Lemma 3 it follows easily that B has an infinite order. The group {A'", B}/ {A 4'", .82} is then a dihedral group, which is not modular. This contradicts· to our assumption that @ is an M-group. It must be therefore a=fJand A'"B=BA'". We now study the structure of M'.'groups which cont.ain elements of infinite orders. Theorem 5. Let @ be an M-group which contains elements of infinite orders. The group i, which consists of all elements of finite orders of @, is then an abelian group.

72

[6]

720

Kenkiti IWASAWA

Proof. Let Z be an element of infinite order and A, B be any two elements of Cf.. AZ is then also of infinite order: otherwise the order of Z would be finite (cf. Theorem 1). If we denote the order of B by m~ it is then from Lemma 3 ZBZ-1

= Y,

(AZ)B(AZ)- 1 = Y',

(rr', m)

= 1.

Consequently

AYA-1 = Y'.

(16)

This shows that Cf. is either abelian or Hamiltonean. If Q: is Hamiltonean it contains a quaternion group 0. By Lemma 3 Z gives an automorphism of finite order in O :

zoz- = o. 1

A suitable power Zo of Z is therefore commutative with every element of D. {Zo, O} / {Zt} is then a 2-group which is not modular (cf. 2 in sl). We thus come to a contradiction and Q: must be abelian. 2. We consider next @/Cf., that is to say, such M-groups, which contain no element of finite order except the identity. We first prove a lemma, which is a special case of the assumption stated in §§1, 2. Lemma 5. An M-group which is generated by two elements of order 2 is a finite group. Proof. Let @ = {A, B}, A2 = E2 = 1. As any element of @ has a finite order according to Theorem 1, put C=AB,

It is then

@

= {A, C} and A -1cA

= c-

1 •

The order of @ is therefore at most 2m. The structure of @/(f, is then given .by the following theorem. Theorem 6. An M-group which contains no element of finite order except the identity is abelian. The proof is rather complicated. We devide it into two parts, the latter of which is available for any M-group. Lemma 6. Let A and B be elements of an M-group which contains no element of finite order except the identity. If {A}r,{B} =t=l, then the subgroup {A, B} is cyclical.

[6]

73 721

On the structure of infinite M-groups.

Lemma 7. Let A and B be elements of infinite orders of an M-group. If {A}A{B} = 1, then A and B are commutative with one another;

AB=BA. Proof of Lemma abelian. Let

6.

Clearly it is sufficient to prove that {A, B} is

{A}r.{B}

=

{Z},

A"= Z

and suppose A 4= BAB- 1• As

A"= (BAB-1)", {A}r.{BA..B-"1} =t= 1, we put

{A}r.{BAB- 1}

= {C},

From Lemma 4 we have a = a'. A 2 = BAB- 1, it holds

Ar=

A;= C,

C ==A"= BA"'B- 1 •

Therefore if we denote A1

{A1}r.{A2}

= A,

= {C}.

As the M-group {Ai. A2}/{C} is generated by two elements of finite orders, it contains no element of infinite order. There is consequently such ,y( =I= 0), that

Let It is then ~ ~ er and from 58,.... er c::: { A 1 } ,.... { A2} = {C } we can readily see that 5i3,....er = {(AzCT)"} = {C1+"T}. As (A1CT)"= Cl+"T it is also 58,....~ = {C 1 +"T}. Now ~l'-'58 contains (A1CT)(A2C·)- 1 = A1Ai1. therefore ~v~ = {A1CT, A2C', er}= {A1, Az} = 58'-'er. We have thus

m: C:::.

(£,

= err.58, m:v58 = erv5S. m: = [, hut this is clearly a contradiction.

~'"'~

From (1) we conclude Remark that 'Y is not zero as A1A21 is of infinite order. We used here essentially the assumption that the group has no element of finite order except the identity. Pro.of of Lemma 7. We put @ = {A, B} and prove that @ is abelian. For that purpose we must first prove the following Lemma 8. Let @ = {A, B} be an M-group as above defined. If @ has an abelian normal subgroup m: with a finite index, then @ itself is abelian.

14

[6]

722

Kenkiti Iw ASA w A

Proof. We remark first that @ contains no element of finite order except the identity : for if Z =t= l is any of such elements we have {A, Z}

= {A,

.&»},

and there exist infinitely many subgroups between {A, Z} and {A} (cf. Lemma 2). But this contradicts to the lattice isomorphism {A, Z}/{A}- {Z}/{A}r.{Z}. Now according to the assumption there is suitable integers that A", B~e2{,

f.l,

fl

such

Hence we have {A}"{B~AB-~}4=1. If A=t=B~AB-~, then we come to a contradiction as in the proof of Lemma 6. We must have therefore B~

or

= A-1B~A.

From {B}"{A- 1BA}=l=l we have again similarly AB=BA.

or We now prove Lemma 7. Denote @={A, B},

@1

=

{ABA- 1, B}, @2

= {A, B2},

@a=

{AB2A-1, B2}.

By Lemma 2 we have (17)

@1

= {A", B}

or

@i

=

{B}.

As ABA - 1 is contained in @1 , A is a normaliser of @1 and @1 is accordingly normal in @. In the same way we can see that @s is normal in @ii. From @a= {(AB)B 2(AB)- 1, B2 } it is also normal in {AB, B2}, consequently in @ = {A, AB, B2}. As @1/@a is generated by two elements of order 2 it is a finite group according to Lemma 5. We first assume that @/@a is infinite. @/@1 is then also infinite and from (17) we have @1 = {B}: otherwise @/@1 would be of order a. It follows ABA- 1 = B~ and by Lemma 4 AB== BA, that is to say, @ is abelian. Next let @/@s be finite. As @a~@2 and @=f:;@2 (cf. Lemma 2), @/@a is not the identical group. As any finite M-group is meta-abelian (cf. §1) we see from this that the commutator group @' of @ does not coincide with @: @=l=@'. If @':={A} and @' c:: {B} it follows then from {A}r-.{B} = 1 that @' = 1 and that @ is abelian. We assume therefore for instance @' ${A}. From {A}=il:i{A} v@' we have then by Lemma 2

{A}v@'={A,B'}, v=pO.

[61

75 On the structure of infinite M-groups.

AB {A}'-'@'J@' is a cyclic_ group generated by u such that mod.

.A,

723

there exists suitable

@'.

B 1 = A"'B-" is then contained in @' and {A}v@'={A, B"}={A, B1}. If {A}f"'\{B1 }=\:=1, then it follows from Lemma 6 that {A}v@' is abelian (cf. the above remark), and as @/{A}v@' is finite./@ is also abelian according to Lemma 8 Let therefore be

AB @' is a suogroup of {A, B1} and contains {B1} it is or according to Lemma 2. Assume first @'={Bi}, By Lemma 4 it is easy to see that B1 is contained in the center of @ and that {A}v@' is abelian. @ itself is then an abelian· group according t°' Lemma 8. We assume therefore @'

= {A",

B}.

is then a finite group. In this case. @' is not abelian : otherwise would be abelian and @' = 1 accordiag to Lemma 8. Now @' have the same structure as @. If we denote the commutator group of @' by @", @' /@" is also finite, but not the identical group. @" /@"' is again finite, where @111 is the commutator group of @". @/@"' is then also a finite M-group. But according to §1, 1-4 any finite M-group is meta-abelian: we have thus met a contradiction. @ must be conseguently abelian in any case. Theorem 6 is now clear from these two lemmas. Let @ be any M-group, which contains elements of infinite orders, and G: be the normal subgroup of @ which consists of all elements of finite order in @. By Theorem 5, 6 (if and @/@ are both a.belian groups. It remains therefore only to investigate how i is extn Lattice-ordered Groups, Proc. Imp. Acad. ·Japan, 18,

(1942). See Nakayama, I.e. H. Nakano, Teilweise geordnete A.lgebra, Jap. Jour. Ma.th. 17 (1941). This paper · is -referred to as N. T. in the following. ( 4) For Lemma 1, 2 see H. Freudental, Teilweise geordnete Moduln. Proc. Amsterdam 39 (1936). . ( 2) ( 3)

81

[7]

82 Kenkichi IWAS\AWA

778

Proof. We have

V a,b, > a"'b"' > a,.b'I therefore

Va,b~2 a,Vb"', On the other hand,

Va.Vb, 2. a.b.,

whence Va~Vb~ 2. Va,b,.

Lemma 2. For any sequence

and arbitrary b of

@,

we have (Vas),.....b

= V(a,r,b).

Proof. As it is sufficient to prove V(a,b- 1)r.l = V(a,b- 1 '"'1), We may assume b = 1. It is then by Lemma 1

\la~= ((Va,)vl)((\/a,)r.1)( 6 )

= (V 1 and h? 1 of ~* and .p9.11 re· .spectively. ·Put h

= V ([m]x). mem

'The join in the right-hand side always exists as x ~ [m]x, and for any

.k' e ~\\JI we have by Lemma 3 hr. I k' J = CV([m]x)) "I k' I = V([m]xr.l k' D= 1 .

'Therefore h is contained in .~,m and h ~ 1 . . If we put further xh- 1 = k, we have k~l and

Im j,.-,k =Im lr-.xh- 1 :::.:;;:.

J

m \r.x([m].~)-1

= 1,

·whence lmlr.k=l,

k

E ~\\11'

which completes the proof. With respect to the projector [p] it holds almost every theorems in N. T.; for instance [p]ab

=

[p]a[p]b,

[p](a :::=:: b)

=

[p]a;;::: [p]b.

More generally we have [p] Va= V[p]a,

if the join Va exists. Theoremes 3.5, 3.6, 3.7, 3.8, 3.14, 3.15, 3.16, 3.17, ( 9)

Cf. S. Bochner and S. R. Phillips, Additive set functions and vector lattices; Ann.

of Math. 42 (1941).

[7]

86 Kenkichi !WA.SA.WA.

782

3.18, 3.19, 3.20, 3.21, 3.22, 3.23; 3.24, 3.25 in N.T. are also valid, but we omit the proofs, as those given in ~-T~ can be easily transfered to our· non-abelian case. with minor modifications. § 2. We now prov~ some more lemmas which are ·to be used afterwards. Lemma 6. Let a~ 1 . Then. any a: , such that 1 ~ x ~ ti" (n ; any natural number), can be descrihed as product of elements y, with 1 sy, s a. Proof. Induction with respect to n. Put y = x°(xl"\a- 1)· 1, then

Therefore 1.~ysa,

·By the assumption of the induction (x;....an· 1) is products of such y', that: 1 y' a, which completes the proof. Lemma 7. No element of @ except the identity has a finite order. Proof. Let am= l and put b =·(11"\a). It is then for any natural number n b"' = (I '""al"\ • • • ,-,a") .

s s

Therefore there are only finite 'different elements among the powers of b. b is consequently also of a finite order, lf == 1. b is obviously < l, but: if b 1 , then bH 1 , which contradicts to 1f· 1 = b- 1 1 . It must betherefore b = 1. in the same way we have b' = (1 va) == 1; whence ·a= 1. Lemma 8. In @ aba'" 1 r,.b = 1 implies b = 1 . Proof. As .Pb is a normal subgroup @ , b and aba·1 are both contained in ~b. But by the assumption aba- 1 e ~,. therefore aba· 1 e •P.bl""\S?b = 1 , b = 1 . Lemma 9.. In @ aba- 1 = b- 1 implies b = 1 . · Proof. We have a(l vb)a- 1 = (1 vb- 1) == b., . that is ab+a· 1 = b.. As. b, r,.b_ = 1, it follows from Lemma 8 that b. = b. = 1, b = 1. Lemma 10. In @ a2 = b2 implies a = b . Proof. Put

>


1 , b > l . From 00

[p1]a

00

= m=l V(Pt"a),

[p:i:Jb

= ·n-l V(py·"b)

we see in virtue of Lemma 13 that [p2]a and [p?]b are commutative. If we denote by 6 . the set of all singular elements in @ , then we have by Theorem 2 @

= .~:; X ·'P.~ •

But as .~ 6 is generated by VJp]x, (x > 1), it follows from what is Pe·.:: mentioned just above, that ,P:; is an abelian group. It is also evident that St'..; is a conditionally complete lattice-group, which has no singular element except the identity. 4. To prove that @ is abelian it is therefore sufficient to prove that St'.:; is an abelian . group. We assume accordingly for a time that @ = Sl'e, that is to say, that @ contains no singular element except that identity. Lemma 14. Let a::>1, a>·x2 , x2::_l. Then there exists an element x'. such that x' x , a ::?. x~ 2 • Proof. We prove first that if a>l, there is an element x', satisfying a> x'2 , x' 1. As a is not singular by the assumption, there is an element a', such that a:> a'> 1, x' = a' "aa1 - 1 :=p 1. It follows then x'2 < aa1 - 1a1 = a, Now let x2 and put a' = x- 2a 1 . From what is just mentioned above there is such z, that a'~z2 , z>L Put xzx- 1 "z=JJ, x' = xy . Then we have y ::.,> 1 by Lemma 8 and it holds

> >

a>

>

which completes the proof. Lemma 15. If a> 1, there exists an element x such that a= x2 • Proof. We construct a sequence,

[7]

89 785

On the Structure o/ Condi,tiomdly Complete Lattice-Gi·oups X1=

l

1'.>0

from (3) it follows by simple calculations (ei,.e; 1) ... = [e>.] a ([e.,.]at 1 ~ (e,.e; 1)1' for l

{ 5)

>·µ,.

~

[e,.e; 1] a

Now given any natural numbers m, n, · we devide equally the .interval [O, m] into

1e

= 2"m subintervijls of length 0

i,i :

= ~ < 11 < ...... < ,l~ = ~ ;,

l~ == ;,. •

From (5) we have then

and whence

It follows ~-

~

1

1 ~ A s;\.[em]a ~ I\ eJ;' .. ~1

n-1

~

= 1 , n-1 V Sm,,. = [em]a ,

co

and as Ve.,,.= p by (4), we have finally( 13 ) m•l

""

..

V sm. .. = m•l V [em]a - [p]a =a.

m,tz.-1

(12) Cf. the pr.oof of Satz 5.15 in N.T. (13) Cf; N.T. Satz S.23.

91

[7] On tke Stri£Cture of Conditumally Com'f)let~ Lattice-Groups

787

If we take another b ~ 1 in !QP , we have_ similarly

where

s,,.,..

is like s,,.," a product of powers of some parts of p . s... " and

Sm,,"'' are therefore commutative with one another and so are a and b consequently. As op,, is generated by such elements a, and b, • • ·, we have thus proved that .'Qi, is an abelian group. Now @/R:i, is :isomorphic

to .~P according to Theorem 1 ; the· commutator group @' of @ is therefore contained in RP. But it is easy to see that the meet of all R,,, p~l, is the identical group. We have thus cw·= 1, that is to say, that @ is abelian. Theorem 3. Any conditionally complete lattice-group is abelian. § 5. We now investigate further the structure of conditionally complete lattice-groups which are just proved to be abelian. We return to the decomposition @ = RG X .'Q$ ' where 0 is the ·set of all singular elements in @. It is almost evident from the consideration in § 4, that R6 is a coI11.plete vector la.ttice, the structure of which has been studied by many authors.. We have therefore only to consider ~ 9 • Now, by Lorenzen, Clifiord and Nakayama( 14 ), ,p,:; can be imbedded lattice-group-isomorphically in a direct product of linear groups S't: We have the representation for a E@ a ___.,. ( •••••• , a't , .••••• ), a't E 2't .

We may assume of course that 2't consists of such a't, which corresponds to some a in @. We· denote by X2 · the set of all coordinates -r such that s't = 1 for every s in 6 , and by X1 · the set of the resting coordinates. If we denote further by m, (i = 1, 2) the direct product of .8.. , TeX, (i = f, 2), .\'6 is imbedded isomoi'phically in 9'lJ X % and We have

a - - (a1 , a2), We prove then that the lattice-homomorphic mapping a-a1 is an isomorphism. Suppose ao - - (1, nf ), (14) P. Lorenzen, Abstrakte BegrUndung der multiplika1;iven Idealtheorie, Math. Zeitschr. 45 (1939>; ':A. H.,Clifford, Partially ordered ·abelian groups, Ann. Math. 41 (1940,) Cf. also Nakayama, 1. c. ·

[71

92 Kenkichi Iw ASA w A

788

then I ao I -

As

(1,

I nf I) .

s - - (s1, 1)

= 1, [s] laol = 1 , whence ao=l . laol = .es V[8]laol = 1,

for every s E ·Z , we have laol r,.s

We may therefore assume from the beginning that X2 is an empty !!et and that for any coordinate .,. there is. some s in 0 such that 81: 1 . For such s1:, assume that there is some a"' satisfying·

>

81:

Putting

'/a"'? '· 1 .

a'= (avl)r.s, we have a~== a"'.

But as s is a singular element it holds a' r-.sa1 - 1 = 1, and whence = a,: l"'\S~a"' - 1 = 1, which contradicts to the. linearity of 21:, Th:us each 2.. , having a minimal element s,., is isom.orphic to the lattice~ group of all rational integers. . Now we call two· coordinates T1, 'Tz are associated when there is no element a in @ such .that a ..t, > 1, a.1:2 1. It then easy .to see .that there is also. no ~lement a in @ such that a.; 1 = 1 , a"'~ 1 and· ,that all coordinates can be thus divided int.o classes of mutually associated ones. We denote these classes by C£. and select from each (£~ a representative 'T•. It is then clear that the -correspondence a~l"\s1:a~- 1

is

>

a ___,.· ( ••.•••• , a-:\1 , •••••• ) gives an isomorphi~ mapping of @ into the direct pr-oduct of 21:v. If we now take two different coordinates .,."' , 'Tp. , then there exists by definition an element a in @ such that a1:v 1, a.,."' =1. Substituting a by a\Jl if necessary, we may suppose that a> 1 . Form then the meet a* = j\a·,. where a is selected as above corresponding to .each G:1:.,. with 'T• ==l= r11-. The meet exists always according to a> I , and as a..v 1 arid ~"' is cyclic, it follows a~ 1 and a~. .,. = 1 for .,.~=I=.,..,.. We have thus proved that ·P?i is a restricted direct product of groups which are isomorphic to the group of all rational integers ordered as usual and that any conditionally complete lattice-group @ is the direct produ~t of such a group with a vector l_at.tice. It is also c~ear that such a decomposition of @ into the direct product is unique, for _in a vector

>

~

>

>

[7]

93 On ths Structurs of C(}'ll,ditionally Complets Lattics·Groups

789

lattice there is no singular element except the identity, whereas any restricted direct product of cyclic groups is generated by its singular elements. We have thus proved the following final theorem: Theorem 4. Any conditionally complete lattice-group is the direct product of two conditionally complete lattice-groups ; one is a vector· lattice and the other is a restricted direct product·. of groups of all rational ·integers.

Mathematical Institute, Faculty of Seience. Tokyo Imperial University.

[8]

Einige Sätze über freie Gruppen

Proc. Imp. Acad. Japan, 19-6 {1943}, 272-274.

Mathematisches Institut, ·Kaiserliche Universität zu Tokyo. (Comm. by T. TAKAGI, M.r.A., Ji:me 12, 1948.)

1. Es sei@ eine beliebige Gruppe. Die durch 31 =@, ,8,.=[@, 3.,._Jl) (n=2, 3, ... ) definierte Kette von Untergruppen von @

.81 :::, .82 :::, 3a :::, ·.. :::, ß,. :::, " · heisst die absteigende Zentralreihe von @2J. 3,-d3i ist dann im Zentrum von @/3, enthalten, also abelsch und hat endlich viele Erzeugende, ( 1)

wenn@ selbst endlich viele Erzeugende besitzt. Wenn ,8„ für irgendein n mit der Einheitsgruppe zusammenfällt, so nennt man @ nilpotent3>. Es gilt dann · Satz 1. @ sei eine nilpotente Gruppe mit endlich vielen Erzeugmden. D(1f' Durchschnitt aller Normalteiler. mit endlichen Indizen ist dann die Einheitsgruppe. Zum Beweis schicken wir einen Hilfssatz voraus. Hilfssatz. Es sei @ eine beliebige Gruppe und 'lc ein Normalteiler von @ mit endlich vielen Erzeugenden. Ist "[@, ,Je] eine endliche Gruppe itnd im Zentrum O von @ enthalten, so enthält ,Je eine Untergruppe sm mit einem endlichen Index [,Je : S:ITT], die in O enthalten ist und durch ·endlich viele Elemente erzeugt 'IJ,Yird. Beweis. l sei die Ordnung von [@, W]. Da [@, ,Je] in 5· enthalten ist, so gilt fur beliebige A e @, Ne m (A, Ni= (A, N') =1°.

Daher ist N' ein Element aus O und Wl= {N1 ; Ne ,Je} hat alle behaupteten Eigenschaften. Beweis von Satz 1. Wir setzen

@=,Bi :::,,fü:::, ... :::, ß,.=1 i=2, ... ,n.

iJ.,, ij,

(i1 < i,z

< ... < ik)

seien sämtliche unendliche Gruppen unter ij.. Der Satz ist offenbar richtig falls k=O. Wir wenden also die Induktion nach k an. Sei zunächst ik ~ n. Nach dem obigen Hilfssatz, angewandt auf 2,

...,

~'k

0, there is a F(s) E 3' such that t'n 一 t,.�O, Ft,.=F,,."Gn. p(E,Gn)�s n

なる如きIn, t'n伍が存在する. t,., tがは[a, b]に

と全く同様であるから特に述べる必要もあるまい. §3. 先づ次の定理を證明する.

高) d(tn(u)))

定理1は無限に分解可能な法則の意味から考へれば 極めて営然の結果と云へるが次の如き意味に於て若千

riabl,8





8/

註6), P. LeYy, p. 180参照 註1), 伊藤氏の論丈の補題3. 1参照

[ 9 ]

100

2 1 9

確率法則の一径薮集合について一—岩澤

の興味が存する. @ Pち §2 I て於て述べた如く 2 の線

, . 若し凡ての n=l,2,・・・・ iこ蜀し tn~ がとすれば t

型部分集合に於てほ距離 p による位相を順序による

は輩調増大であるから

l i m t n = t ' '

位相によつて置換へることか出末るから線型・ 径敷集 合なる概念は p を用ひずに順序だけによつて定義ナ

"7ao

が存在するがこれから

ることが出来る.よ.つて上の定理 1によれば無限に分 解可能な法則類は順序だけにより特徹付けられる.類 の概念も叉順序こにより定義されたのであるからこ れによつて無限に分解可能な法則自身順序だけによつ で特徽付け得ることがわかろのである. §4. €:i= { F t ; tE1 } を任意の一径敷半群とせよ.

RItE を含むから適嘗に座標をづらせることにより 一般性を失ふことなく ( 1 3 ) 尻,=F と恨定しても差支へない.

lim 凡~.. = F , 1 1 ・

n→ .

F舛 : E であるからこれは不合理である四 e i ' 、l : l :径敷 tの大小によつて線型に順序附けら 故に l れた可換な Archimedes的半群でしかも順序の意味 で連結されてゐるからそれほ正の賓敷全堕のつくる半

'上で定義され ( 0 ,o o )な 群と同型である.ょつて I る謳間の凡ての値をとる輩調増大函敷 ( 1 9 ) s : g ( t ) が存在してこれによ

t ,1 を Iに含まれる任意の賓敷とし ( 1 4 )

とすれば t "は t ,t ' により一意的に定まるが tを一

( 1 5 )

t " = J e ( t ' )

としよう.椴定によりこれは t ' の連績函敷である. 叉定職から宜ちに

( 1 6 )

J : ( t ' ) = か( t ) .

さて Iは正敷を含むものと恨定してその一つを t>O とせよ.然らば任意のが E Iに蜀し常に ( 1 7 )

/ i ヽ ( t ' ) > t '

となる.何となれば f c ( t 1 ) . ; ; ; ; t 1 なろ如き t ' が存在し

(O}=t>O で且 fヽは連績であるから たとすれば fヽ f i ( t " )=t " を滴足する t "が存在する.即ち ( 1 8 )

( 2 0 )

凡*応1=Fs+•'

凡*凡,=Ft"

定しておけばそれはがの函敷であるからこれを

F 1 *凡n=F,,,

これから容易に F,=E なることが知られる,.g pち

t>Oで Fc=F0 であるからこれは G の定義に矛盾す る.よつて常に ( 1 7 )が 成 立 す る 換 言 す れ ば 凡 * 砂 =凡"且 t>O であれば t">t'なることがわかつた. 蓋間 Iの非負なる部分を I ' とし R'={Fe;tE判 とすれば上の考察により特にR'も半群であることが

o n v o l u t i o nにより径数が輩調 知られるがそこでは c

b径敷を愛換すれば

®'={瓦;〇~sOに謝し

皿={和;

O~s'~s}

とおけば ! 1 1 1 ,は 明 か に 丑 と 凡 と を 結 ぶ 線 型 一 径 敷 集合であるから定理 lにより凡は無限に分解可能で

' sの特性函敷を c p , ( z ) とすれば ( 5 )に ある.ょって P より

( 2 1 )

r p , ( z ) = e x p{ , J , , ( z ) } , , i , , ( O ) = O

とおくことが出来る. ( 2 0 )により

( 2 2 )

c p , ( z ) c p , , ( z )=r p 1 + r { z )

であるから ( 2 1 ) を代入して

和( z ) + , t , , や )= , t , 1 + 1 1 ( z )

( 2 8 ) , t , , ( z )は

8

について連綾であるから ( 2 8 )から直ちに

,t,,(z)=s• 中. ( z ) .

( 2 4 )

を得る.即ち cph)=c p ( z ) とおけば

( 2 5 )

c p s ( z ) =r p ( z ) ' , 。 ; ; ; ; ;sO とし F eの n個の積

て考へよう.

を凡*れ*・ ・・ ・*F,=Fen とするとき n を十分大きく

a ) Iが唯一貼 0から成る場合.

' 1 : : .野し t n > t ' となる.何となれば とれぽ,任意の t 9)

註6 )の P .Levr,p .9 1 ,定 理 2 9 ,I参 照

1 0 ) 例へば L evy-Doeblinの補助定理 ( P .L e v y ,p . 8 ) を用ぴればよい. 1 5 5 ,定理 4

101

[ 9 ) 日本敷學物理學會誌

220

第十七巻

第六鱗

このときは明かに 6={E} で問題はない

定されたが遥に ( 2 7 )乃至 ( 3 1 }で典へられるもが一径

b l Iが正敷を含み負敷を含まぬ場合.

敷半群であることは明白であるから次の定理が成立す

§いこ述べた考察により逹嘗に径敷を嬰換すれば

る .

( 2 7 )

6={凡;〇~s O を一つとり ( 1 5 ) の函敷 f i . ( t ' ) を考察する. J e . ( O )=t,>0であり且 f凶')は t ' に闊し蓮綬である ' が負で十分小であれば f 叫' ) > 0 となる. こ から t ' の一つを t , とし fヽ , C t ) を考へる. の様な t f t , ( O ) = = t 1 0 f t 1 ( t 2 ) = 0 , O( G) and bounded continuous representations of G in the following sense : i) For a given continuous representation x(g)-T(x) of £(G) is of course given by the norm JlxH in .(3),

114

[11]

115 68

[Vol. 20,

K; lWASAWA.

(4)

T(x)= tx(g)D(g)µ(dg) 5>

for all x(g) in V 1• 11>(G). We denote this representation of G by Di{a). ii) Conversely, if a-D(a) is an arbitrary bounded continuous representation of G and if we define T(x) by (4), the mapping x(g)-T(x) · gives a continuous representation for x(g) in £1l. 17>(G). We denote this representation by Tr:l..x). iii) i) and ii) give mutually inverse correspondences: if D1 =DT1, then T1=TDi and if T2=TD2 , then D2=DTa· · iv) Equivalent. representations correspond to each other: if ATi(x)A-1 =Ta(x), then ADT,(a)A-1 =DT2(a) and if BD1(a)B-1 =D2(a), then BTDkx)B- 1=TDh;). From Theorem 1 follow immediately some corollaries: Let a~D(a) be a bounded measurable representation of G. If we· put

T(x)=J x(g)D(g)µ(dg), G

x(g)-'> T(x) gives, as before, a continuous representation of vi. 11>(G). By Theorem 1 we have thus T(x) = TD,(x) with some bounded continuous representation Di(a) of G and hence D(a)=Di(a). Thus Theorem 2. Any bounded measurable representation of G is

continuous. In a similar way we obtain by a simple calculation the following Theorem 3. If G is locally compact, but not compact, then there is no representation of G belonging to £P(G) (p > 1) except the zero representation, which maps every element of G to the zero matrix. On the other hand if G is compact, any representation belonging to £'P(G) (p > 1) is bounded and continuous. Now, as a bounded representation of G is always completely reducible, it follows from Theorem 1, iv) tha,t a continuous representation of L'1·P>(G) is completely reducible. But the converse js also true. It holds namely Theorem 4. A representation T(x) of L' 1• 11>(G) is continuous if and only if it is ·completely reducible. Especially a irreducible representation of L'1•. 11>(G) is .always continuous. This theorem is equivalent to the following 1,1>>(G), such that Theorem 5. Let M be a two-sided ideal of the rest class ring £CL 1>> / M is of finite. dimension. . £f M is then semi-simple if and only if M is closed in L'1•P>( G). Especially a maximal ideal M is always closed in vi. P>(G). These theorems can be regarded as a generalization of the complete reducibility of the g:roup ring of a finite group. The above mentioned relation between ideals and representations of V 1• P>(G) is explicitly given by Theorem 6. Let {D} be a class of equivalent bounded continuous

v

-----~·--··--------··-----··-. - ----- - - - - - - - - · · - - - 5) The right-hand side means a matlix with (i,j)-component where D(a)= {dij(a)}.

)JCU)di {g)µ(dg), 3

116

[11)

No. 2.J

On Group Rings of Topological Groups.

69

irreducible representations of G and D(a) be a representant of it. Then all functions x(g) of £< 1• v>( G) satisfying Jax(g)D(g)µ(dg)==O,

constitute a maximal two-sided ideal M{D} in V 1·v)(G). {D} -+M{D} thus gives a one-to-one correspondence between all classes of equivalent bounded continuous irreducible representations of G and all maximal two-sided ideals M of £(G), for which V 1·P)/M is of finite dimension. If we define the classes of (not necessarily irreducible) representations of G suitably, then the result of Theorem 6 can be extended to those classes of representations of G and all closed two-sided ideals M of V 1• P>(G), for which V 1• v)/ M is of finite dimension. It follows then also, that for V 1• vl( G), p = 1, 2, . . . the ideals of that kind correspond one-to-one to each other. § 3. In order to establish corresponding theorems for Segal's group ring J?i- 1• P> (G), we have only to prove the following Lemma. Let M be a two-sided ideal in £ 0,P>(G) such that the rest class ring V 1• v)/ M is of finite dimension· and has a unit element. Then a two-sided ideal M' of H- 1• 11 \G) can be uniquely determined, so that it holds M=M' r.£0,P)(G)'

R. § 4. We now extend our Theorem 1 to representations of G and V 1• P>( G) by bounded operators in a Hilbert space ~- Let B be the ----•--•W'••·---

6) We define the norm in R.

On the other hand we call a representation x(g)-T(x)(T(x)eB) of V 1· P)( G) in B " proper ", if T (x )! = O(f e .p) for all x(g) e £( G) 10>• 111 T(x) Ills C 1 \Ix!\, We can now prove the following Theorem 8. There is a one-to-one correspondence between continuous proper representations of £< 1,P>(G) in B and bounded measur~ ablem proper representations of G in B in the following sense : i) For such a representation x(g)-> T(x) of £< 1· Pl(G) in B, there is a bounded measurable proper representation a->D(a) of G, so that (6)

T(x)= )ox(g)D(g),u(dg) 121

for all x(g) in J) 1· P>(G). Such D(a) is uniquely determined by T(x). ii) Conversely, if a-> D(a) is such a representation of G, T(x) in (6) gives a continuous proper representation of £(G) in B. iii) Above correspondences are mutually inverse. iv) Equivalent representations correspond to each other 13J. If the measure ,u is not only left invariant, but also right invariant, then we can obtain some more precise results. We can thus prove for example the following theorem. Theorem 9. If G has an invariant Haar measure, then measurable unitary representations of G in B are all strongly continuous14>. Detailed proofs of above theorems will appear elsewhere. They need some considerations on non-commutative normed rings15> and rings of operators in a Hilbert space, as will be also discussed there precisely16 >. 9) Ill A Ill means the bound of the operator A. 10) Thus we consider in B the uniform topology. 11) That is to say, that (D(g}f,f') is µ-measurable for any f,f' in ~-

12) (6) means (T(x}f,/')=) i;(g) (D(gif.f')µ(clg) for any f,f' in ~13) Cf. Theorem I, iii), iv). 14) Cf. K. Kodaira: Uber die Gruppen der messbaren Abbildungen, Proc. 17

(1941), 18-23. 15) Some of the theorems, obtained by I. Gelfand, concerning commutative normed rings can-be transferred to our non-commutative case. 16) Cf. also author's .note in Zenkoku Sijo Sugaku Danwakai, 246 (1942), 15221555, 251 (1943), 167-186.

[12]

••

Uber nilpotente topologische Gruppen I Proc. Japan Acad., 21.;.3 (1945), 124-137,

M11the111atisches Institut, Kaiserliche Univel'Bität su · Tokyo. (Comm. by T. TAICAGI, M,1,A,, Maroh 12; 1945.)

In der abstruten Gruppentheorie spielen bekanntlich die höheren Kommutntorgruppen, die abstejgenden · bzw, die aufsteigenden Zentralreihen wichtige Rollen, Sie charakterisieren besondere Klaaaen von Gruppen, nimlich auflösbare bzw. nilpotente Gruppen, deren Struktur, vor allem bei endlichen Gruppen, ,eingehend untersucht worden ist1), Dem.entsprechend sollen im folgenden die Komm.utatorgruppen und die Zentralgruppen· auch für topologische Gruppen definiert werden tmd dann die Struktur der auflösbaren l:xl.w; der nilpotenten topologi, -sehen, insbesondere kompakten Gruppen untersucht werden, Es sei @S eine separable topologische Gruppe ?lld Q, -Ql, ...... (nicht not· wendig abgeschlossene) Untergruppen von ®· Wir bezeichnen mit ,e, -Ql,,,,,,. bzw, [9J, -Ql] ,. ,, •• ,die abgeschlossenen Hüllen l:xl.w. die abstrakten Kommutatorgruppen

(im Shm.e der abstrakten Gruppentheorie)2) solcher Untergruppen, wie leicht ersichtlich

[9J,

(1)

Wir nennen

[9J, -Ql]

Es gilt !lann

,Ql] =[fJ, ~].

die topologwche Kom1nuf)J,torgntppe von $l und -Ql,

Nun ersetzten wir die gewöhnlichen Kommuta.torgruppen durch die topologischen ttnd definieren die topologischen höheren Kommutato:rgruppen $f(QS) bzw. die topoiogiflChen absteigenden Zentralgruppen Sf(QS) folgendermassen, Es sei

$f(@)=Qs, St(®)=CI und für jede Ordnungsmhl '1J mit 'l'J$:C«s)!l'/m schon für 'TJ~(@Jml, ;l)~(®tlns::[;l)~(@)ltm, ~~(@lm1mJ s::[$~(@), 2:l~(@}],t/,t=$f(QJ)!?/!?.

setzen wir voraus dass

wenn

eeine Limeszahl ist, so. gilt

${(@/!?)= A~;(@/m;)~ /\ (~:{@),t/m;)>( A~~{@)),t/,t ~ ...... . Von einem gewissen m- an ist also dim. S'j,,,=dim; Sl.. +1=, ..... · und die l·Kotnponente2' von $.2,.., S'j... +1, .. , ... fallen zusammen, Wenn man diese Gruppe mit ~o bezeichnet; so sind Sl../~01 S'im+1/~0, ...... sämtlich kompakt und diskret, also. endliche Gruppen und es gilt · $.2,,J.~0>$.2,,.+1/~o> •.... ,, Daher gibt es ein n-(~m) mit ~.;/,to=Sln+1/~0=··••

00 ,

also $l,.=S'j,.+1=······,

w.z. b.w. Nunzcigen wir $!(")=S)!+1(@)= .... ,. EsgenÜgtoffenbarnur S)!(@)= $!+1(®) nachzuweisen, Wir setzen· voraus $!(®)~$!+1(@) und wählen ein Element.g, !nit g e ~!(@), g $ $!+!(@). Da @/$!+i®) eine kompakte Grup· pe ist; gibt es eine stetige Darstellung D von@/$!+1(CI); welche g nicht auf die Einheitsmatrix E abbildet. Es ist also (6) D($!(®))~{E}, Aus der Kompaktheit von

«s und aus$!(®)=/\. '$,f((I} sieht man

D($!(@))= AD($!(")).

"

Da es aber

D($t(®))> D($f(~))~D($f(Gi$))> · · · ·. ist und D($f(C$))=D(@) eine kompakte Liesche Gruppeis( so folgt nach

Hilfssatz 3 =D($!(@)). D($! (QS}) = D( $!+1(®)) = ,.. ··· Andererseits ist aber $!(@)/D!+.1(@) eine abelsche Gruppe. . E,s ist also ,nach Hilfssatz 1, (4') und (7) (7)

D($!(@})=D($!+1(®})=-$!+1W(@))={$!(D(@)), ~.~(D(@))] :;:=.[D($!(@)), D($!(@))] =[D(S)!(@)), D($!(@))] ~D($!+i(@))= {E} , was mit (6) in Widerspruch steht. 'Es muss also $!(@)=$!+1(@) sein, 1) Im folgenden .soll I immer solche Gruppe bedeuten,

2) Die zusammenhändgende Komponente, welche das Einselement enthält, bekanntlich· einen Normalteiler,

Sie bildet

(12]

121 Ober nilpotentc topologische GrupJl1ffl, I.

No. 3.]

127

Man beweist nun$!(@/~)=$!(@)~/~ folgendermassen:

$!~@/,t)= /\ $;':(@/,t)= /\ $!(@)~/~ (nach (4')) tt ff =( /\ $!'(@)),t/~ (nach der Kompaktheit von @S) . .

~

= $!{@),t/~ •

Gcnan so zeigt man S!(@/m) = S!(@lmt~ und damit ist der Hilfssatz

2 in allen Teilen bewiesen. Nun definicrm. wir auflösbare bzw,• nilpotente topologische· Gruppen,

@S

heisst topologisch aujlösba1· wenn für irgendeine Ordnungszahl~ $f(@)=e ist, 1)

und topologisch nach iinten bzw. nach oben nilpotent wenn fÜr irgendeine Ordnung8zahl ~ Stl@)=e l:YLw, U(@)=@ gilt.

Gilt zwar $f(@)=e (S[(@);:::e

oder lt(@)=@), $:~e cs:(@)~e oder %:(@)~@) für

'IJ$1(@)>$!(@)>......

~$...(@)=81

/\ ,., $,.(@) = $ ..(~)=8.

Nach ljilfssatz 3 gilt also von einem n an ~ ..(@)=~..+1(@)= •. ,

= ~ ..(@)=6,

@ ist mithin eine zusammenhlfügende, kompakte, und· im Sinne der Lieschen

Theorie auflösbare Gmppe, ist also bekanntlich abelsch; Für eine nach unten nilpotenten Gn-.ppe können wir noch mehr sagen,

Es

gilt nämlich Sa.tz 5,

Die l·Komponeut @0 einer kompakten nacili UJlten nilpotenten

Grµppe @ ist im Zentrum von·@ enthalten, Beweis. annehmen,

Genau wie im vorigen Beweis dürfen wir @ als Liesche Gruppe Da @ offeabar auflösbar ist, so ist @0 nach Satz 4 eine zusammen·

hängende kompakte abelsche Liesche Gruppe, also· direktes Produkt e11dlichvieler

Gruppen st., ,velche der mod, 1 reduzierten additiven Gruppe

ft aller, reellen

Zahlen isomorph sind

1\ .

@0 =st1

x !,x ftsX , .. xft~, !,~! (i=l, 2, ... , m),

Die Elemente von @0 können also durch "Koordinaten" dargest,ellt werden:

(8)

h,"".. (h1, h,, •.• , hin)

0"4, mod. 1),

Die Transformation zha:- durch ein Element 1

ID

aus @ iJ;i.duziert -dann in @0_

einen Automorphismus, ,velcher durch eine ganz24hlige Matrix A(a:) dargestellt wird:

(9) rohx- 1++(h1 h9,; ..... , h,;.)A(a:). Aus (8), (9) ergibt sfoh (10) (z, h)=a:hx-1h-1++(h 1,.hs, ...... , n...) (.A(x)-E) und daraus durch n-malige Kommutatorbildung

(a:, (x, ..• (x, h)) ••• )+·•{h1, hii,···, h,..) (A{x)--E)", Man sieht .aber genau so wie im. Beweis des Sat.zes 4, dass für genÜgend grosses n I) Vgl. L. Pontrjagin, Topologioal groups (1939), S. 170.

123

[12) Uber uilpotente topologische Gruppen, I.

No. 3.J

.ß,.(@)=e stattfindet.

129

Für- beliebige k; a: ist clahn der n-te Kommutator

(a:, (z, ..• (a:, h), .. ) immer gleich dem Einselement u~d daraus folgt

(A(x)-E)"=O. Die Eigenwerten von A(x) sind mithin sämtlich gleich 1.

Da aber x-+A(x)

offenbar eine Darstellung der endlichen Gruppe ®/@ 0 gibt, so schliesst man sofort

A(z)=E,

:ih,a:-1:=h,

also

Wir untersüchen nun. die Gruppe @/@o, d,h, 0-dimensionale kompakte ·nach unten nilpotente Gruppe, Eine 0-dimensionale kompakte Gruppe Ja.eisst eine p-Gruppe, wenn es der Limes einer @S„-adischen Reihe 'Von endlichen p·Gr)lppen ist, Eine Untergruppe 9) einer 0-dimensionalen kompakten Gruppe ® heisst eine p-Sylowgruppe von

®, wenn· 9) eine maximale p-G:ruppe in ® ist. Unter diesen Begriffsbildungen hat v. Dantzig die Sylowsä,tze in der Theorie der endlichen Gruppen auf die -topolog).schen Gruppen übertragen.1> Er hat z, B, bewiesen, dass alle p-Sylow-

Für die rw.ch u1iten nilpotenten Grup-

_,gruppen von·@ efnarder konjugiert sind.

pen beweisen wir nun folgende Sätze über Sylowgl'uppen.

Hillasatz 4,

Wen:Q. @ nach unten nilpotent ist, so ist

das· direkte Pro-

dukt seiner Sylowgl'l;l:ti)en, Beweis,

Es sei 9) eine p.Sylowgruppo von @S und

.

/\,tto=6

,t1>,ti_>,ts>•·•, .

(11)

eine Reihe offener (und zugleich abgeschlossener) Normalteiler von®, welche ein Umgebungssystem in· ® bestimmt.

Es lässt sich dann leicht ~ige.n, dass

9) ,t'1,t, eine p-8ylowgl"llppe der endlichen Gruppe @5/,t, .ist und dass man umgekehrt jede p-Sylowgruppe -von® in solcher Weise als. (1) 11-adj.schen times der

@/,t, erhiilt. Setzt man nun @S als nach unten nilpotent voraus, so sind @Jg/:, sämtlich endliche nilpotentc Gruppen, also direkte Produkte ihrer Sylowgmppen. Daraus und aus dem oben Gesagten folgt sofort., dass. ®

p-Sylowgruppen -von

selbst direktes Produkt der Sylowgttippen i..'lt, Hilfssatz 5, .Beweis.

Jede p-Oruppe ist naoh unten nilpotent•

Es sei 9) eine p-Orlippe und

,t, seien wie .oben in

_gebllngsfilYstem bildende, otfene Normalteiler von

1WR.~ immer endliche p-Gruppen. ß..;(tl~,,t,

vlfll

Nach Definition sind·.

Daraus folgt.

Ci=1,2, ... ), a1so S.,(S,)=e.

Satz 2 lµld Hilfssätze 4, 5 können 1) D.

9).

(11) ein Um-

wu in·fo]gender,u Satz zusammenfassen,

Dantzig, Zur t'op.>logisclien Algebra IU. Comp. Math. Vol. 3 (1936).

[12]

124 180

K. lWAsAWA.

Satz 6,

Dafür, dass eine O·dimensionale kompakte Gruppe nach untl:!n

nilpotent sei, ist. notwendig und hinreichend, dass •sie direktes Produkt ihrer Sylowgruppen ist, Nach diesem Satz reduziert sich die Untersuchung derStrukten der O·dimen· sionalcn kompakten nach unten nilpotenten Gruppe in einem gewissen Sinne auf clic der Struktur der endlichen p-Gruppen, Bei dieser Gelegenheit sei noch darauf aufmerksam gemacht, dass sich ver·· S% ..(@)@,, >@0 und aus der Endlich·

128

[12]

134

[Vol. 21,

~. l'ivASAWA.

für eine gewisse natii.rliche Zahl

n. %..(@)=V%.,(@) ergibt dann

'

%,.(@)@o=@. Nun ist aber@/%,,(@)~ @J%,.(@)

I""'\

@0 nach Satz 8 abelsh, somit. %,.+ 1(@)=@.

Wir zeigen nun, dass es immer eine kompakte nach oben nilpatente Liesche Gmppe gibt, welche die Klasse

e=ru+1i, n=l, 2, ... besitzt.

ft wie vorher die Gruppe rler reellen Zahlen motl, 1, und dede!nige .;\_utomorphismus von S'e, welcher jedes x von st mit seinem Inverse - x vertm1·. Es sei

seht.

(1'

Wir setwn N-•fBI ,.., l)·,..., ,..2_,, nn•n-1-'-•"'-( "") II!}3~, "'J w,vw - . . , -- - . . , ,

-,

,•e 3~• BI

J!."" HU ;v

Es zeigt sich leicht tl, h, dass @ die Klasse ru

+1 besitzt,

Dieses Beispiel zeigt zugleich dass eine kom·

pakte n.a(:h oben nilpotente Grüppe nicht immer nach unten nilpat.ent ist, , a1)pe", i,.=l

irreduzible Formen in k [ a1 , b], wo µ bzw. v alle verschiedene Konjugierten von f;' ~w. (~~

i;) in bezug auf k(b) bzw, k(a 1) durchläuft und p•",p•1 die Ex·

ponenten von k(b, l;')/k(b) und k(a 1 ,

l, -;,)/lc(a1)

sind.

Man beweist dann

genau wie oben, dass F1(a1, b) mit G(a 1 , b) bis auf einen konstanten Faktor zusammenfällt,

k(b,

!;', "1') ist aber,

wie schon bemerkt, ttber k(b) separabel.

Der Exponent p•" von k(b, f;')/k(b) ist folglich gleich 1. in diesem Fall mit k(b, !;', 1i1 ) zusammen,

Fern-r fällt k(b,

e

1)

Zum Beweis setze man

n*= [k(b, !;', ,;')_: k(b, ,')], also n'=n*n", a, Es gibt dann n* Schnittpunkte von O mit Eine durch

(e, ,;')

Hb, die erste Koordinate

e'

haben,

hindurchgehende allgemeine Hyperfläche vom Grad (s', t')

1) Vgl. Chow, 1. c, 2) Vgl. v. d, Waetden, 1. c. 3), S. 146. 3) n" war der reduzierttl Grad von k(b, ~')fk (b), ab)e Erweitemng ist, so ist es gleich [k(b, k (b)].

n:

Da aber k (b, E')/k (b) eine separ--

[13]

135

216

K. IWASAWA,

[Val. 21,

(e, r/)

in keinem Punkt mit der ersten

schneidet aber die Kurve O ausser iu

e;x -l;~x1=0 in Pi,"' schneidetOnur

Koordinate/;' mehr, denn die Byperebene

0

in endlichvielen Punkten, falls C nicht von der Form ( 0 x 02 ist.

n*= 1 und folglich k(b, e')=lc(b, f;', r,')

sein und es gilt

"' F1(a1, b)= IIJi(l;'~P.\ a1),

(3)

n'= [k(b, !;', 1J'): k(b)],

µ=1

Wir betrachten nun den Exponent p"1• Korrespondenzen haben

Es muss also

Nach der Theorie der algebraischen

$, ij, a, isomorphe algebraische Relationen tiber k wie

t;*,'IJ*, at, die man erhält, wem1 man durch einen allgemeinen Punkt (!;*, 'IJ*) '\'On O eine allgemeinste Hyperfläche f 1(x, af) vom Grads legt1l, Ji(x, a1)=a1o~+a11X~-\c1+a12~- 2-:c~+ ......

Setzt man

=a1ox:;+ J;(x, all, a12, ...... ), so lassen sir.h

atv

z.B. so bestimmen, dass für unbestimmte

*

U10= gesetzt wird2',

(4)

fi(l;*,

ati, ~l!i, ..... .

ati, at, ...... ) f;f'

Es gilt dann

k(a1, ~ 17}::::::: k(at, f;*, 'YJ* )= k(I;*, r,*, atifafo, a'f:i/ato ... ... ) =k(l;*, 'TJ*, a'tii/a?i, a"ts/afi, ...... )

und ähnlich

(5)

k(a1, $")~k(at, l;*)=k(l;*, a'fülafi, at.ifafi, ...... ),

Da I;*, 'TJ * und af,;/afi, afs!afi; ...... von einander algebraisch nnabhängig sind,

so folgt aus (4), (5) dasa der Grad und Exponent von k(a, ~, 'ii)/k(a 1 , 1;) de~nigen von Tc(!;*, r/)/k(e'") glekh sind, Nun ist~ als Schnittpunkt der auf P 1 projizierenden Bildkurve von O mit der allgemeh)en Hyperflache /1(x, a1):;::.:0 in P 1 sicher separabel tiber k(a 1). p' 1 ist daher gleich dem Exponent von k(a1, !, ij)/k(a1, f), also nach oben denjenigen von Tc(~*, rJ*)/k(I;*).

Setzt 1nan alsdann

ri= [k(~*, r,*): k(~*)] = [k(a,'

e, 11): k(a1, e)J =r1p'1,

sofälltr1 mit dem reduzierten Grad vonlc(~*,17*)/k(~*) zusammen, da k(~*) ein algebraischer Funktionenkörper von einer Veränderlichen ist3).

(6)

Daraus folgt auch

n1=r1[Tc(a1, l):k(ai)],

Nun defin:eren wir die Multiplizität eines Schnittpunktes einer irreduziblen Kurve O mit einer allgemeinerHyperfläche f(x,y,a)=O vom Grad (s,t),st>O folgendermassen: es ist gleich 1) Vgl, z.B. v,d. Waerden, 1. c, 3), Kap. V. 2) Man kann natürlich ~o*O annehmen. 3) Vgl. z.B. M. Deuring, Arithmetische Theorie der Korrespondenzen algebraischer Funktionenkörper, I, Crelle's.Jout. 177 (1937).

136

[13]

No. 4.)

i)

Der Bezoutsche Satz in z eifach proje'..tiven Räume ,

21.T

I,fjir8t>O,

ii)

dem Expenent p• 1 von k(e*, 'IJ*)/k(e*), für t=O,

iii)

dem Exponent p 02 von k(e*, 1J*)/k(1J*), für 8=0,

wobei ( e*, 1J*) ein allgemeinen Punkt von C über k bedeutet.

Es sei bemerkt;dass die so definierte Multiplizitli.t .allenfalls nur von O abhängig ist,

Nach· dieser Pefinition kann man den oben Bewiesenen in folgender einfacher Form zusammenfassen, ~atz I.

Es seien j(x,

y, a), g(x, y, b)

allgemeine ll'ormen von Graden

(s, t), (I, t!) tmd O eine irreduzible Kürve in P,,.,,,., den Hyperßächen

Schnittpunkte von O mit

J(x, y, a)=O, g(x, y, b)=O seien, mit Multiplizitäten ver-

sehen, (e·), .,,0 ·)), i=I, ...... , a, (e'.

Die Schnitt;

punkte von 01 mit einer allgemeinen Hyperfläche f 1(x, a 1)=0 in P, seien

~P>, g=[k( sind und von diesen Sclmittpnnkten fallen je Jl' einander zusammen.·

Im folgenden bikeichnen wir die Gesamtheit der niit Vielfaehheiten versehenen Schnittpunkte von O mit f(x, y, a)=O immer mit

[C,f(x, y, a)], Nun seien JCx, y, a), Ji(x, y, a 1), Graden (8, t), (81,

f 2(x, y, ae)

allgemeine Formen von

t,), (s;, t2} tmd es gelte

. s=s1+82,

t=t,+~, it>O, 81t1~0, ~ts>·O.

Es gilt dann der Satz 4,

Setzt man

J(x, y, a*)= J1(x, y, a1)J2(x, Y, a2), so geht [O,J(x, y, a)] bei dor relationstreuen Spezialisierung a ~ a* in die Summe von [C,J1(x, y, a1)] und

[9,J2(x, 'Y; ~)].

1) Genau dann ist ja k (f, 71)/k (~) eine endliche Erweitei:ung.

137

[13]

[Vo!. 21,

K. lWASAWA,

218

Wir nehmen zunäe.hst an, dass O wirklich J1(x,

Beweis.

y, ai)=O und

f2(x, y, a2)=0 schneidet und setzen [O,J(x,y,a)]= ((f!;CP),?JC?)),p=L ......... , a), [O,J1(x,y,a1)]={(1!;1,?J''")),q=l, ...... ,ßJ, [0,flx, y, a2)] {(/!;"'... \ ?J"(-.)), T=1, , ..... , 'Y},

=

Setzt man ferner

=

g(x, y, b)

l m

I 2 b;,jXiYh [O, g(x, y, b)] =

•-Oj=O

{( /!;*''"\ 11*("')), w=l, , ..... , o},

so gilt nach Satz 1, ( 7)

F(a, b)=IIg(f!;m, ?Jm, b)=IIJ(f!;*C"'\ 11*